Commit | Line | Data |
---|---|---|
b8aa1682 JD |
1 | /* |
2 | * Copyright (C) 2012 - Julien Desfossez <jdesfossez@efficios.com> | |
3 | * David Goulet <dgoulet@efficios.com> | |
cd60b05a | 4 | * 2013 - Jérémie Galarneau <jeremie.galarneau@efficios.com> |
7591bab1 | 5 | * 2015 - Mathieu Desnoyers <mathieu.desnoyers@efficios.com> |
b8aa1682 JD |
6 | * |
7 | * This program is free software; you can redistribute it and/or modify | |
8 | * it under the terms of the GNU General Public License, version 2 only, | |
9 | * as published by the Free Software Foundation. | |
10 | * | |
11 | * This program is distributed in the hope that it will be useful, but WITHOUT | |
12 | * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or | |
13 | * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for | |
14 | * more details. | |
15 | * | |
16 | * You should have received a copy of the GNU General Public License along | |
17 | * with this program; if not, write to the Free Software Foundation, Inc., | |
18 | * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA. | |
19 | */ | |
20 | ||
6c1c0768 | 21 | #define _LGPL_SOURCE |
b8aa1682 JD |
22 | #include <getopt.h> |
23 | #include <grp.h> | |
24 | #include <limits.h> | |
25 | #include <pthread.h> | |
26 | #include <signal.h> | |
27 | #include <stdio.h> | |
28 | #include <stdlib.h> | |
29 | #include <string.h> | |
30 | #include <sys/mman.h> | |
31 | #include <sys/mount.h> | |
32 | #include <sys/resource.h> | |
33 | #include <sys/socket.h> | |
34 | #include <sys/stat.h> | |
35 | #include <sys/types.h> | |
36 | #include <sys/wait.h> | |
896010e3 | 37 | #include <sys/resource.h> |
173af62f | 38 | #include <inttypes.h> |
b8aa1682 JD |
39 | #include <urcu/futex.h> |
40 | #include <urcu/uatomic.h> | |
70626904 | 41 | #include <urcu/rculist.h> |
b8aa1682 JD |
42 | #include <unistd.h> |
43 | #include <fcntl.h> | |
f8be1183 | 44 | #include <strings.h> |
896010e3 | 45 | #include <ctype.h> |
b8aa1682 JD |
46 | |
47 | #include <lttng/lttng.h> | |
48 | #include <common/common.h> | |
49 | #include <common/compat/poll.h> | |
50 | #include <common/compat/socket.h> | |
f263b7fd | 51 | #include <common/compat/endian.h> |
e8fa9fb0 | 52 | #include <common/compat/getenv.h> |
b8aa1682 | 53 | #include <common/defaults.h> |
3fd27398 | 54 | #include <common/daemonize.h> |
b8aa1682 JD |
55 | #include <common/futex.h> |
56 | #include <common/sessiond-comm/sessiond-comm.h> | |
57 | #include <common/sessiond-comm/inet.h> | |
b8aa1682 JD |
58 | #include <common/sessiond-comm/relayd.h> |
59 | #include <common/uri.h> | |
a02de639 | 60 | #include <common/utils.h> |
d3ecc550 | 61 | #include <common/align.h> |
f40ef1d5 | 62 | #include <common/config/session-config.h> |
5312a3ed JG |
63 | #include <common/dynamic-buffer.h> |
64 | #include <common/buffer-view.h> | |
70626904 | 65 | #include <common/string-utils/format.h> |
b8aa1682 | 66 | |
2a635488 | 67 | #include "backward-compatibility-group-by.h" |
0f907de1 | 68 | #include "cmd.h" |
2a635488 | 69 | #include "connection.h" |
d3e2ba59 | 70 | #include "ctf-trace.h" |
2a635488 | 71 | #include "health-relayd.h" |
1c20f0e2 | 72 | #include "index.h" |
d3e2ba59 | 73 | #include "live.h" |
2a635488 | 74 | #include "lttng-relayd.h" |
2a174661 | 75 | #include "session.h" |
2a635488 | 76 | #include "sessiond-trace-chunks.h" |
2a174661 | 77 | #include "stream.h" |
f056029c | 78 | #include "tcp_keep_alive.h" |
2a635488 JR |
79 | #include "testpoint.h" |
80 | #include "tracefile-array.h" | |
81 | #include "utils.h" | |
82 | #include "version.h" | |
83 | #include "viewer-stream.h" | |
b8aa1682 | 84 | |
4fc83d94 PP |
85 | static const char *help_msg = |
86 | #ifdef LTTNG_EMBED_HELP | |
87 | #include <lttng-relayd.8.h> | |
88 | #else | |
89 | NULL | |
90 | #endif | |
91 | ; | |
92 | ||
5569b118 JG |
93 | enum relay_connection_status { |
94 | RELAY_CONNECTION_STATUS_OK, | |
a9577b76 | 95 | /* An error occurred while processing an event on the connection. */ |
5569b118 JG |
96 | RELAY_CONNECTION_STATUS_ERROR, |
97 | /* Connection closed/shutdown cleanly. */ | |
98 | RELAY_CONNECTION_STATUS_CLOSED, | |
99 | }; | |
100 | ||
b8aa1682 | 101 | /* command line options */ |
ce9ee1fb | 102 | char *opt_output_path, *opt_working_directory; |
35ab25e5 | 103 | static int opt_daemon, opt_background, opt_print_version, opt_allow_clear = 1; |
a8b66566 | 104 | enum relay_group_output_by opt_group_output_by = RELAYD_GROUP_OUTPUT_BY_UNKNOWN; |
3fd27398 MD |
105 | |
106 | /* | |
107 | * We need to wait for listener and live listener threads, as well as | |
108 | * health check thread, before being ready to signal readiness. | |
109 | */ | |
110 | #define NR_LTTNG_RELAY_READY 3 | |
111 | static int lttng_relay_ready = NR_LTTNG_RELAY_READY; | |
0848dba7 MD |
112 | |
113 | /* Size of receive buffer. */ | |
114 | #define RECV_DATA_BUFFER_SIZE 65536 | |
115 | ||
3fd27398 MD |
116 | static int recv_child_signal; /* Set to 1 when a SIGUSR1 signal is received. */ |
117 | static pid_t child_ppid; /* Internal parent PID use with daemonize. */ | |
118 | ||
095a4ae5 MD |
119 | static struct lttng_uri *control_uri; |
120 | static struct lttng_uri *data_uri; | |
d3e2ba59 | 121 | static struct lttng_uri *live_uri; |
b8aa1682 JD |
122 | |
123 | const char *progname; | |
b8aa1682 | 124 | |
65931c8b | 125 | const char *tracing_group_name = DEFAULT_TRACING_GROUP; |
cd60b05a JG |
126 | static int tracing_group_name_override; |
127 | ||
128 | const char * const config_section_name = "relayd"; | |
65931c8b | 129 | |
b8aa1682 JD |
130 | /* |
131 | * Quit pipe for all threads. This permits a single cancellation point | |
132 | * for all threads when receiving an event on the pipe. | |
133 | */ | |
0b242f62 | 134 | int thread_quit_pipe[2] = { -1, -1 }; |
b8aa1682 JD |
135 | |
136 | /* | |
137 | * This pipe is used to inform the worker thread that a command is queued and | |
138 | * ready to be processed. | |
139 | */ | |
58eb9381 | 140 | static int relay_conn_pipe[2] = { -1, -1 }; |
b8aa1682 | 141 | |
26c9d55e | 142 | /* Shared between threads */ |
b8aa1682 JD |
143 | static int dispatch_thread_exit; |
144 | ||
145 | static pthread_t listener_thread; | |
146 | static pthread_t dispatcher_thread; | |
147 | static pthread_t worker_thread; | |
65931c8b | 148 | static pthread_t health_thread; |
b8aa1682 | 149 | |
7591bab1 MD |
150 | /* |
151 | * last_relay_stream_id_lock protects last_relay_stream_id increment | |
152 | * atomicity on 32-bit architectures. | |
153 | */ | |
154 | static pthread_mutex_t last_relay_stream_id_lock = PTHREAD_MUTEX_INITIALIZER; | |
095a4ae5 | 155 | static uint64_t last_relay_stream_id; |
b8aa1682 JD |
156 | |
157 | /* | |
158 | * Relay command queue. | |
159 | * | |
160 | * The relay_thread_listener and relay_thread_dispatcher communicate with this | |
161 | * queue. | |
162 | */ | |
58eb9381 | 163 | static struct relay_conn_queue relay_conn_queue; |
b8aa1682 | 164 | |
896010e3 JG |
165 | /* Cap of file desriptors to be in simultaneous use by the relay daemon. */ |
166 | static unsigned int lttng_opt_fd_cap; | |
167 | ||
d3e2ba59 JD |
168 | /* Global relay stream hash table. */ |
169 | struct lttng_ht *relay_streams_ht; | |
170 | ||
92c6ca54 DG |
171 | /* Global relay viewer stream hash table. */ |
172 | struct lttng_ht *viewer_streams_ht; | |
173 | ||
7591bab1 MD |
174 | /* Global relay sessions hash table. */ |
175 | struct lttng_ht *sessions_ht; | |
0a6518b0 | 176 | |
55706a7d | 177 | /* Relayd health monitoring */ |
eea7556c | 178 | struct health_app *health_relayd; |
55706a7d | 179 | |
23c8ff50 JG |
180 | struct sessiond_trace_chunk_registry *sessiond_trace_chunk_registry; |
181 | ||
cd60b05a JG |
182 | static struct option long_options[] = { |
183 | { "control-port", 1, 0, 'C', }, | |
184 | { "data-port", 1, 0, 'D', }, | |
8d5c808e | 185 | { "live-port", 1, 0, 'L', }, |
cd60b05a | 186 | { "daemonize", 0, 0, 'd', }, |
b5218ffb | 187 | { "background", 0, 0, 'b', }, |
cd60b05a | 188 | { "group", 1, 0, 'g', }, |
896010e3 | 189 | { "fd-cap", 1, 0, '\0', }, |
cd60b05a JG |
190 | { "help", 0, 0, 'h', }, |
191 | { "output", 1, 0, 'o', }, | |
192 | { "verbose", 0, 0, 'v', }, | |
193 | { "config", 1, 0, 'f' }, | |
3a904098 | 194 | { "version", 0, 0, 'V' }, |
ce9ee1fb | 195 | { "working-directory", 1, 0, 'w', }, |
a8b66566 JR |
196 | { "group-output-by-session", 0, 0, 's', }, |
197 | { "group-output-by-host", 0, 0, 'p', }, | |
35ab25e5 | 198 | { "disallow-clear", 0, 0, 'x' }, |
cd60b05a JG |
199 | { NULL, 0, 0, 0, }, |
200 | }; | |
201 | ||
3a904098 | 202 | static const char *config_ignore_options[] = { "help", "config", "version" }; |
cd60b05a | 203 | |
a3bc3918 JR |
204 | static void print_version(void) { |
205 | fprintf(stdout, "%s\n", VERSION); | |
206 | } | |
207 | ||
208 | static void relayd_config_log(void) | |
209 | { | |
210 | DBG("LTTng-relayd " VERSION " - " VERSION_NAME "%s%s", | |
211 | GIT_VERSION[0] == '\0' ? "" : " - " GIT_VERSION, | |
212 | EXTRA_VERSION_NAME[0] == '\0' ? "" : " - " EXTRA_VERSION_NAME); | |
213 | if (EXTRA_VERSION_DESCRIPTION[0] != '\0') { | |
214 | DBG("LTTng-relayd extra version description:\n\t" EXTRA_VERSION_DESCRIPTION "\n"); | |
215 | } | |
7f5ed73a JR |
216 | if (EXTRA_VERSION_PATCHES[0] != '\0') { |
217 | DBG("LTTng-relayd extra patches:\n\t" EXTRA_VERSION_PATCHES "\n"); | |
218 | } | |
a3bc3918 JR |
219 | } |
220 | ||
cd60b05a JG |
221 | /* |
222 | * Take an option from the getopt output and set it in the right variable to be | |
223 | * used later. | |
224 | * | |
225 | * Return 0 on success else a negative value. | |
226 | */ | |
7591bab1 | 227 | static int set_option(int opt, const char *arg, const char *optname) |
b8aa1682 | 228 | { |
cd60b05a JG |
229 | int ret; |
230 | ||
231 | switch (opt) { | |
232 | case 0: | |
896010e3 JG |
233 | if (!strcmp(optname, "fd-cap")) { |
234 | unsigned long v; | |
235 | ||
236 | errno = 0; | |
237 | v = strtoul(arg, NULL, 0); | |
238 | if (errno != 0 || !isdigit(arg[0])) { | |
239 | ERR("Wrong value in --fd-cap parameter: %s", | |
240 | arg); | |
241 | ret = -1; | |
242 | goto end; | |
243 | } | |
244 | if (v < DEFAULT_RELAYD_MINIMAL_FD_CAP) { | |
245 | ERR("File descriptor cap must be set to at least %d", | |
246 | DEFAULT_RELAYD_MINIMAL_FD_CAP); | |
247 | } | |
248 | if (v >= UINT_MAX) { | |
249 | ERR("File descriptor cap overflow in --fd-cap parameter: %s", | |
250 | arg); | |
251 | ret = -1; | |
252 | goto end; | |
253 | } | |
254 | lttng_opt_fd_cap = (unsigned int) v; | |
255 | DBG3("File descriptor cap set to %u", lttng_opt_fd_cap); | |
256 | } else { | |
257 | fprintf(stderr, "unknown option %s", optname); | |
258 | if (arg) { | |
259 | fprintf(stderr, " with arg %s\n", arg); | |
260 | } | |
cd60b05a JG |
261 | } |
262 | break; | |
263 | case 'C': | |
e8fa9fb0 MD |
264 | if (lttng_is_setuid_setgid()) { |
265 | WARN("Getting '%s' argument from setuid/setgid binary refused for security reasons.", | |
266 | "-C, --control-port"); | |
267 | } else { | |
268 | ret = uri_parse(arg, &control_uri); | |
269 | if (ret < 0) { | |
270 | ERR("Invalid control URI specified"); | |
271 | goto end; | |
272 | } | |
273 | if (control_uri->port == 0) { | |
274 | control_uri->port = DEFAULT_NETWORK_CONTROL_PORT; | |
275 | } | |
cd60b05a JG |
276 | } |
277 | break; | |
278 | case 'D': | |
e8fa9fb0 MD |
279 | if (lttng_is_setuid_setgid()) { |
280 | WARN("Getting '%s' argument from setuid/setgid binary refused for security reasons.", | |
281 | "-D, -data-port"); | |
282 | } else { | |
283 | ret = uri_parse(arg, &data_uri); | |
284 | if (ret < 0) { | |
285 | ERR("Invalid data URI specified"); | |
286 | goto end; | |
287 | } | |
288 | if (data_uri->port == 0) { | |
289 | data_uri->port = DEFAULT_NETWORK_DATA_PORT; | |
290 | } | |
cd60b05a JG |
291 | } |
292 | break; | |
8d5c808e | 293 | case 'L': |
e8fa9fb0 MD |
294 | if (lttng_is_setuid_setgid()) { |
295 | WARN("Getting '%s' argument from setuid/setgid binary refused for security reasons.", | |
296 | "-L, -live-port"); | |
297 | } else { | |
298 | ret = uri_parse(arg, &live_uri); | |
299 | if (ret < 0) { | |
300 | ERR("Invalid live URI specified"); | |
301 | goto end; | |
302 | } | |
303 | if (live_uri->port == 0) { | |
304 | live_uri->port = DEFAULT_NETWORK_VIEWER_PORT; | |
305 | } | |
8d5c808e AM |
306 | } |
307 | break; | |
cd60b05a JG |
308 | case 'd': |
309 | opt_daemon = 1; | |
310 | break; | |
b5218ffb MD |
311 | case 'b': |
312 | opt_background = 1; | |
313 | break; | |
cd60b05a | 314 | case 'g': |
e8fa9fb0 MD |
315 | if (lttng_is_setuid_setgid()) { |
316 | WARN("Getting '%s' argument from setuid/setgid binary refused for security reasons.", | |
317 | "-g, --group"); | |
318 | } else { | |
319 | tracing_group_name = strdup(arg); | |
320 | if (tracing_group_name == NULL) { | |
321 | ret = -errno; | |
322 | PERROR("strdup"); | |
323 | goto end; | |
324 | } | |
325 | tracing_group_name_override = 1; | |
330a40bb | 326 | } |
cd60b05a JG |
327 | break; |
328 | case 'h': | |
4fc83d94 | 329 | ret = utils_show_help(8, "lttng-relayd", help_msg); |
655b5cc1 | 330 | if (ret) { |
4fc83d94 | 331 | ERR("Cannot show --help for `lttng-relayd`"); |
655b5cc1 PP |
332 | perror("exec"); |
333 | } | |
cd60b05a | 334 | exit(EXIT_FAILURE); |
3a904098 | 335 | case 'V': |
a3bc3918 JR |
336 | opt_print_version = 1; |
337 | break; | |
cd60b05a | 338 | case 'o': |
e8fa9fb0 MD |
339 | if (lttng_is_setuid_setgid()) { |
340 | WARN("Getting '%s' argument from setuid/setgid binary refused for security reasons.", | |
341 | "-o, --output"); | |
342 | } else { | |
343 | ret = asprintf(&opt_output_path, "%s", arg); | |
344 | if (ret < 0) { | |
345 | ret = -errno; | |
346 | PERROR("asprintf opt_output_path"); | |
347 | goto end; | |
348 | } | |
cd60b05a JG |
349 | } |
350 | break; | |
ce9ee1fb JR |
351 | case 'w': |
352 | if (lttng_is_setuid_setgid()) { | |
353 | WARN("Getting '%s' argument from setuid/setgid binary refused for security reasons.", | |
354 | "-w, --working-directory"); | |
355 | } else { | |
356 | ret = asprintf(&opt_working_directory, "%s", arg); | |
357 | if (ret < 0) { | |
358 | ret = -errno; | |
359 | PERROR("asprintf opt_working_directory"); | |
360 | goto end; | |
361 | } | |
362 | } | |
363 | break; | |
364 | ||
cd60b05a JG |
365 | case 'v': |
366 | /* Verbose level can increase using multiple -v */ | |
367 | if (arg) { | |
368 | lttng_opt_verbose = config_parse_value(arg); | |
369 | } else { | |
849e5b7b DG |
370 | /* Only 3 level of verbosity (-vvv). */ |
371 | if (lttng_opt_verbose < 3) { | |
372 | lttng_opt_verbose += 1; | |
373 | } | |
cd60b05a JG |
374 | } |
375 | break; | |
a8b66566 JR |
376 | case 's': |
377 | if (opt_group_output_by != RELAYD_GROUP_OUTPUT_BY_UNKNOWN) { | |
378 | ERR("Cannot set --group-output-by-session, another --group-output-by argument is present"); | |
379 | exit(EXIT_FAILURE); | |
380 | } | |
381 | opt_group_output_by = RELAYD_GROUP_OUTPUT_BY_SESSION; | |
382 | break; | |
383 | case 'p': | |
384 | if (opt_group_output_by != RELAYD_GROUP_OUTPUT_BY_UNKNOWN) { | |
385 | ERR("Cannot set --group-output-by-host, another --group-output-by argument is present"); | |
386 | exit(EXIT_FAILURE); | |
387 | } | |
388 | opt_group_output_by = RELAYD_GROUP_OUTPUT_BY_HOST; | |
389 | break; | |
35ab25e5 MD |
390 | case 'x': |
391 | /* Disallow clear */ | |
392 | opt_allow_clear = 0; | |
393 | break; | |
cd60b05a JG |
394 | default: |
395 | /* Unknown option or other error. | |
396 | * Error is printed by getopt, just return */ | |
397 | ret = -1; | |
398 | goto end; | |
399 | } | |
400 | ||
401 | /* All good. */ | |
402 | ret = 0; | |
403 | ||
404 | end: | |
405 | return ret; | |
406 | } | |
407 | ||
408 | /* | |
409 | * config_entry_handler_cb used to handle options read from a config file. | |
f40ef1d5 | 410 | * See config_entry_handler_cb comment in common/config/session-config.h for the |
cd60b05a JG |
411 | * return value conventions. |
412 | */ | |
7591bab1 | 413 | static int config_entry_handler(const struct config_entry *entry, void *unused) |
cd60b05a JG |
414 | { |
415 | int ret = 0, i; | |
416 | ||
417 | if (!entry || !entry->name || !entry->value) { | |
418 | ret = -EINVAL; | |
419 | goto end; | |
420 | } | |
421 | ||
422 | /* Check if the option is to be ignored */ | |
423 | for (i = 0; i < sizeof(config_ignore_options) / sizeof(char *); i++) { | |
424 | if (!strcmp(entry->name, config_ignore_options[i])) { | |
425 | goto end; | |
426 | } | |
427 | } | |
428 | ||
429 | for (i = 0; i < (sizeof(long_options) / sizeof(struct option)) - 1; i++) { | |
430 | /* Ignore if entry name is not fully matched. */ | |
431 | if (strcmp(entry->name, long_options[i].name)) { | |
432 | continue; | |
433 | } | |
434 | ||
435 | /* | |
7591bab1 MD |
436 | * If the option takes no argument on the command line, |
437 | * we have to check if the value is "true". We support | |
438 | * non-zero numeric values, true, on and yes. | |
cd60b05a JG |
439 | */ |
440 | if (!long_options[i].has_arg) { | |
441 | ret = config_parse_value(entry->value); | |
442 | if (ret <= 0) { | |
443 | if (ret) { | |
444 | WARN("Invalid configuration value \"%s\" for option %s", | |
445 | entry->value, entry->name); | |
446 | } | |
447 | /* False, skip boolean config option. */ | |
448 | goto end; | |
449 | } | |
450 | } | |
451 | ||
452 | ret = set_option(long_options[i].val, entry->value, entry->name); | |
453 | goto end; | |
454 | } | |
455 | ||
456 | WARN("Unrecognized option \"%s\" in daemon configuration file.", | |
457 | entry->name); | |
458 | ||
459 | end: | |
460 | return ret; | |
461 | } | |
462 | ||
2a10de3b JR |
463 | static int parse_env_options(void) |
464 | { | |
465 | int ret = 0; | |
466 | char *value = NULL; | |
467 | ||
468 | value = lttng_secure_getenv(DEFAULT_LTTNG_RELAYD_WORKING_DIRECTORY_ENV); | |
469 | if (value) { | |
470 | opt_working_directory = strdup(value); | |
471 | if (!opt_working_directory) { | |
472 | ERR("Failed to allocate working directory string (\"%s\")", | |
473 | value); | |
474 | ret = -1; | |
475 | } | |
476 | } | |
477 | return ret; | |
478 | } | |
479 | ||
7591bab1 | 480 | static int set_options(int argc, char **argv) |
cd60b05a | 481 | { |
178a0557 | 482 | int c, ret = 0, option_index = 0, retval = 0; |
cd60b05a JG |
483 | int orig_optopt = optopt, orig_optind = optind; |
484 | char *default_address, *optstring; | |
485 | const char *config_path = NULL; | |
486 | ||
487 | optstring = utils_generate_optstring(long_options, | |
488 | sizeof(long_options) / sizeof(struct option)); | |
489 | if (!optstring) { | |
178a0557 | 490 | retval = -ENOMEM; |
cd60b05a JG |
491 | goto exit; |
492 | } | |
493 | ||
494 | /* Check for the --config option */ | |
495 | ||
496 | while ((c = getopt_long(argc, argv, optstring, long_options, | |
497 | &option_index)) != -1) { | |
498 | if (c == '?') { | |
178a0557 | 499 | retval = -EINVAL; |
cd60b05a JG |
500 | goto exit; |
501 | } else if (c != 'f') { | |
502 | continue; | |
503 | } | |
504 | ||
e8fa9fb0 MD |
505 | if (lttng_is_setuid_setgid()) { |
506 | WARN("Getting '%s' argument from setuid/setgid binary refused for security reasons.", | |
507 | "-f, --config"); | |
508 | } else { | |
509 | config_path = utils_expand_path(optarg); | |
510 | if (!config_path) { | |
511 | ERR("Failed to resolve path: %s", optarg); | |
512 | } | |
cd60b05a JG |
513 | } |
514 | } | |
515 | ||
516 | ret = config_get_section_entries(config_path, config_section_name, | |
517 | config_entry_handler, NULL); | |
518 | if (ret) { | |
519 | if (ret > 0) { | |
520 | ERR("Invalid configuration option at line %i", ret); | |
cd60b05a | 521 | } |
178a0557 | 522 | retval = -1; |
cd60b05a JG |
523 | goto exit; |
524 | } | |
b8aa1682 | 525 | |
cd60b05a JG |
526 | /* Reset getopt's global state */ |
527 | optopt = orig_optopt; | |
528 | optind = orig_optind; | |
b8aa1682 | 529 | while (1) { |
cd60b05a | 530 | c = getopt_long(argc, argv, optstring, long_options, &option_index); |
b8aa1682 JD |
531 | if (c == -1) { |
532 | break; | |
533 | } | |
534 | ||
cd60b05a JG |
535 | ret = set_option(c, optarg, long_options[option_index].name); |
536 | if (ret < 0) { | |
178a0557 | 537 | retval = -1; |
b8aa1682 JD |
538 | goto exit; |
539 | } | |
540 | } | |
541 | ||
542 | /* assign default values */ | |
543 | if (control_uri == NULL) { | |
fa91dc52 MD |
544 | ret = asprintf(&default_address, |
545 | "tcp://" DEFAULT_NETWORK_CONTROL_BIND_ADDRESS ":%d", | |
546 | DEFAULT_NETWORK_CONTROL_PORT); | |
b8aa1682 JD |
547 | if (ret < 0) { |
548 | PERROR("asprintf default data address"); | |
178a0557 | 549 | retval = -1; |
b8aa1682 JD |
550 | goto exit; |
551 | } | |
552 | ||
553 | ret = uri_parse(default_address, &control_uri); | |
554 | free(default_address); | |
555 | if (ret < 0) { | |
556 | ERR("Invalid control URI specified"); | |
178a0557 | 557 | retval = -1; |
b8aa1682 JD |
558 | goto exit; |
559 | } | |
560 | } | |
561 | if (data_uri == NULL) { | |
fa91dc52 MD |
562 | ret = asprintf(&default_address, |
563 | "tcp://" DEFAULT_NETWORK_DATA_BIND_ADDRESS ":%d", | |
564 | DEFAULT_NETWORK_DATA_PORT); | |
b8aa1682 JD |
565 | if (ret < 0) { |
566 | PERROR("asprintf default data address"); | |
178a0557 | 567 | retval = -1; |
b8aa1682 JD |
568 | goto exit; |
569 | } | |
570 | ||
571 | ret = uri_parse(default_address, &data_uri); | |
572 | free(default_address); | |
573 | if (ret < 0) { | |
574 | ERR("Invalid data URI specified"); | |
178a0557 | 575 | retval = -1; |
b8aa1682 JD |
576 | goto exit; |
577 | } | |
578 | } | |
d3e2ba59 | 579 | if (live_uri == NULL) { |
fa91dc52 MD |
580 | ret = asprintf(&default_address, |
581 | "tcp://" DEFAULT_NETWORK_VIEWER_BIND_ADDRESS ":%d", | |
582 | DEFAULT_NETWORK_VIEWER_PORT); | |
d3e2ba59 JD |
583 | if (ret < 0) { |
584 | PERROR("asprintf default viewer control address"); | |
178a0557 | 585 | retval = -1; |
d3e2ba59 JD |
586 | goto exit; |
587 | } | |
588 | ||
589 | ret = uri_parse(default_address, &live_uri); | |
590 | free(default_address); | |
591 | if (ret < 0) { | |
592 | ERR("Invalid viewer control URI specified"); | |
178a0557 | 593 | retval = -1; |
d3e2ba59 JD |
594 | goto exit; |
595 | } | |
596 | } | |
896010e3 JG |
597 | if (lttng_opt_fd_cap == 0) { |
598 | int ret; | |
599 | struct rlimit rlimit; | |
600 | ||
601 | ret = getrlimit(RLIMIT_NOFILE, &rlimit); | |
602 | if (ret) { | |
603 | PERROR("Failed to get file descriptor limit"); | |
604 | retval = -1; | |
605 | } | |
606 | ||
607 | lttng_opt_fd_cap = rlimit.rlim_cur; | |
608 | } | |
b8aa1682 | 609 | |
a8b66566 JR |
610 | if (opt_group_output_by == RELAYD_GROUP_OUTPUT_BY_UNKNOWN) { |
611 | opt_group_output_by = RELAYD_GROUP_OUTPUT_BY_HOST; | |
612 | } | |
35ab25e5 MD |
613 | if (opt_allow_clear) { |
614 | /* Check if env variable exists. */ | |
615 | const char *value = lttng_secure_getenv(DEFAULT_LTTNG_RELAYD_DISALLOW_CLEAR_ENV); | |
616 | if (value) { | |
617 | ret = config_parse_value(value); | |
618 | if (ret < 0) { | |
619 | ERR("Invalid value for %s specified", DEFAULT_LTTNG_RELAYD_DISALLOW_CLEAR_ENV); | |
620 | retval = -1; | |
621 | goto exit; | |
622 | } | |
623 | opt_allow_clear = !ret; | |
624 | } | |
625 | } | |
a8b66566 | 626 | |
b8aa1682 | 627 | exit: |
cd60b05a | 628 | free(optstring); |
178a0557 | 629 | return retval; |
b8aa1682 JD |
630 | } |
631 | ||
7591bab1 MD |
632 | static void print_global_objects(void) |
633 | { | |
634 | rcu_register_thread(); | |
635 | ||
636 | print_viewer_streams(); | |
637 | print_relay_streams(); | |
638 | print_sessions(); | |
639 | ||
640 | rcu_unregister_thread(); | |
641 | } | |
642 | ||
b8aa1682 JD |
643 | /* |
644 | * Cleanup the daemon | |
645 | */ | |
7591bab1 | 646 | static void relayd_cleanup(void) |
b8aa1682 | 647 | { |
7591bab1 MD |
648 | print_global_objects(); |
649 | ||
b8aa1682 JD |
650 | DBG("Cleaning up"); |
651 | ||
178a0557 MD |
652 | if (viewer_streams_ht) |
653 | lttng_ht_destroy(viewer_streams_ht); | |
654 | if (relay_streams_ht) | |
655 | lttng_ht_destroy(relay_streams_ht); | |
7591bab1 MD |
656 | if (sessions_ht) |
657 | lttng_ht_destroy(sessions_ht); | |
178a0557 | 658 | |
095a4ae5 | 659 | free(opt_output_path); |
ce9ee1fb | 660 | free(opt_working_directory); |
095a4ae5 | 661 | |
a02de639 CB |
662 | /* Close thread quit pipes */ |
663 | utils_close_pipe(thread_quit_pipe); | |
664 | ||
710c1f73 DG |
665 | uri_free(control_uri); |
666 | uri_free(data_uri); | |
8d5c808e | 667 | /* Live URI is freed in the live thread. */ |
cd60b05a JG |
668 | |
669 | if (tracing_group_name_override) { | |
670 | free((void *) tracing_group_name); | |
671 | } | |
b8aa1682 JD |
672 | } |
673 | ||
674 | /* | |
675 | * Write to writable pipe used to notify a thread. | |
676 | */ | |
7591bab1 | 677 | static int notify_thread_pipe(int wpipe) |
b8aa1682 | 678 | { |
6cd525e8 | 679 | ssize_t ret; |
b8aa1682 | 680 | |
6cd525e8 MD |
681 | ret = lttng_write(wpipe, "!", 1); |
682 | if (ret < 1) { | |
b8aa1682 | 683 | PERROR("write poll pipe"); |
b4aacfdc | 684 | goto end; |
b8aa1682 | 685 | } |
b4aacfdc MD |
686 | ret = 0; |
687 | end: | |
b8aa1682 JD |
688 | return ret; |
689 | } | |
690 | ||
7591bab1 | 691 | static int notify_health_quit_pipe(int *pipe) |
65931c8b | 692 | { |
6cd525e8 | 693 | ssize_t ret; |
65931c8b | 694 | |
6cd525e8 MD |
695 | ret = lttng_write(pipe[1], "4", 1); |
696 | if (ret < 1) { | |
65931c8b | 697 | PERROR("write relay health quit"); |
b4aacfdc | 698 | goto end; |
65931c8b | 699 | } |
b4aacfdc MD |
700 | ret = 0; |
701 | end: | |
702 | return ret; | |
65931c8b MD |
703 | } |
704 | ||
b8aa1682 | 705 | /* |
b4aacfdc | 706 | * Stop all relayd and relayd-live threads. |
b8aa1682 | 707 | */ |
b4aacfdc | 708 | int lttng_relay_stop_threads(void) |
b8aa1682 | 709 | { |
b4aacfdc | 710 | int retval = 0; |
b8aa1682 JD |
711 | |
712 | /* Stopping all threads */ | |
713 | DBG("Terminating all threads"); | |
b4aacfdc | 714 | if (notify_thread_pipe(thread_quit_pipe[1])) { |
b8aa1682 | 715 | ERR("write error on thread quit pipe"); |
b4aacfdc | 716 | retval = -1; |
b8aa1682 JD |
717 | } |
718 | ||
b4aacfdc MD |
719 | if (notify_health_quit_pipe(health_quit_pipe)) { |
720 | ERR("write error on health quit pipe"); | |
721 | } | |
65931c8b | 722 | |
b8aa1682 | 723 | /* Dispatch thread */ |
26c9d55e | 724 | CMM_STORE_SHARED(dispatch_thread_exit, 1); |
58eb9381 | 725 | futex_nto1_wake(&relay_conn_queue.futex); |
178a0557 | 726 | |
b4aacfdc | 727 | if (relayd_live_stop()) { |
178a0557 | 728 | ERR("Error stopping live threads"); |
b4aacfdc | 729 | retval = -1; |
178a0557 | 730 | } |
b4aacfdc | 731 | return retval; |
b8aa1682 JD |
732 | } |
733 | ||
734 | /* | |
735 | * Signal handler for the daemon | |
736 | * | |
737 | * Simply stop all worker threads, leaving main() return gracefully after | |
738 | * joining all threads and calling cleanup(). | |
739 | */ | |
7591bab1 | 740 | static void sighandler(int sig) |
b8aa1682 JD |
741 | { |
742 | switch (sig) { | |
b8aa1682 JD |
743 | case SIGINT: |
744 | DBG("SIGINT caught"); | |
b4aacfdc MD |
745 | if (lttng_relay_stop_threads()) { |
746 | ERR("Error stopping threads"); | |
747 | } | |
b8aa1682 JD |
748 | break; |
749 | case SIGTERM: | |
750 | DBG("SIGTERM caught"); | |
b4aacfdc MD |
751 | if (lttng_relay_stop_threads()) { |
752 | ERR("Error stopping threads"); | |
753 | } | |
b8aa1682 | 754 | break; |
3fd27398 MD |
755 | case SIGUSR1: |
756 | CMM_STORE_SHARED(recv_child_signal, 1); | |
757 | break; | |
b8aa1682 JD |
758 | default: |
759 | break; | |
760 | } | |
761 | } | |
762 | ||
763 | /* | |
764 | * Setup signal handler for : | |
765 | * SIGINT, SIGTERM, SIGPIPE | |
766 | */ | |
7591bab1 | 767 | static int set_signal_handler(void) |
b8aa1682 JD |
768 | { |
769 | int ret = 0; | |
770 | struct sigaction sa; | |
771 | sigset_t sigset; | |
772 | ||
773 | if ((ret = sigemptyset(&sigset)) < 0) { | |
774 | PERROR("sigemptyset"); | |
775 | return ret; | |
776 | } | |
777 | ||
b8aa1682 JD |
778 | sa.sa_mask = sigset; |
779 | sa.sa_flags = 0; | |
0072e5e2 MD |
780 | |
781 | sa.sa_handler = sighandler; | |
b8aa1682 JD |
782 | if ((ret = sigaction(SIGTERM, &sa, NULL)) < 0) { |
783 | PERROR("sigaction"); | |
784 | return ret; | |
785 | } | |
786 | ||
787 | if ((ret = sigaction(SIGINT, &sa, NULL)) < 0) { | |
788 | PERROR("sigaction"); | |
789 | return ret; | |
790 | } | |
791 | ||
0072e5e2 | 792 | if ((ret = sigaction(SIGUSR1, &sa, NULL)) < 0) { |
b8aa1682 JD |
793 | PERROR("sigaction"); |
794 | return ret; | |
795 | } | |
796 | ||
0072e5e2 MD |
797 | sa.sa_handler = SIG_IGN; |
798 | if ((ret = sigaction(SIGPIPE, &sa, NULL)) < 0) { | |
3fd27398 MD |
799 | PERROR("sigaction"); |
800 | return ret; | |
801 | } | |
802 | ||
803 | DBG("Signal handler set for SIGTERM, SIGUSR1, SIGPIPE and SIGINT"); | |
b8aa1682 JD |
804 | |
805 | return ret; | |
806 | } | |
807 | ||
3fd27398 MD |
808 | void lttng_relay_notify_ready(void) |
809 | { | |
810 | /* Notify the parent of the fork() process that we are ready. */ | |
811 | if (opt_daemon || opt_background) { | |
812 | if (uatomic_sub_return(<tng_relay_ready, 1) == 0) { | |
813 | kill(child_ppid, SIGUSR1); | |
814 | } | |
815 | } | |
816 | } | |
817 | ||
b8aa1682 JD |
818 | /* |
819 | * Init thread quit pipe. | |
820 | * | |
821 | * Return -1 on error or 0 if all pipes are created. | |
822 | */ | |
7591bab1 | 823 | static int init_thread_quit_pipe(void) |
b8aa1682 | 824 | { |
a02de639 | 825 | int ret; |
b8aa1682 | 826 | |
a02de639 | 827 | ret = utils_create_pipe_cloexec(thread_quit_pipe); |
b8aa1682 | 828 | |
b8aa1682 JD |
829 | return ret; |
830 | } | |
831 | ||
832 | /* | |
833 | * Create a poll set with O_CLOEXEC and add the thread quit pipe to the set. | |
834 | */ | |
7591bab1 | 835 | static int create_thread_poll_set(struct lttng_poll_event *events, int size) |
b8aa1682 JD |
836 | { |
837 | int ret; | |
838 | ||
839 | if (events == NULL || size == 0) { | |
840 | ret = -1; | |
841 | goto error; | |
842 | } | |
843 | ||
844 | ret = lttng_poll_create(events, size, LTTNG_CLOEXEC); | |
845 | if (ret < 0) { | |
846 | goto error; | |
847 | } | |
848 | ||
849 | /* Add quit pipe */ | |
c7759e6a | 850 | ret = lttng_poll_add(events, thread_quit_pipe[0], LPOLLIN | LPOLLERR); |
b8aa1682 JD |
851 | if (ret < 0) { |
852 | goto error; | |
853 | } | |
854 | ||
855 | return 0; | |
856 | ||
857 | error: | |
858 | return ret; | |
859 | } | |
860 | ||
861 | /* | |
862 | * Check if the thread quit pipe was triggered. | |
863 | * | |
864 | * Return 1 if it was triggered else 0; | |
865 | */ | |
7591bab1 | 866 | static int check_thread_quit_pipe(int fd, uint32_t events) |
b8aa1682 JD |
867 | { |
868 | if (fd == thread_quit_pipe[0] && (events & LPOLLIN)) { | |
869 | return 1; | |
870 | } | |
871 | ||
872 | return 0; | |
873 | } | |
874 | ||
875 | /* | |
876 | * Create and init socket from uri. | |
877 | */ | |
7591bab1 | 878 | static struct lttcomm_sock *relay_socket_create(struct lttng_uri *uri) |
b8aa1682 JD |
879 | { |
880 | int ret; | |
881 | struct lttcomm_sock *sock = NULL; | |
882 | ||
883 | sock = lttcomm_alloc_sock_from_uri(uri); | |
884 | if (sock == NULL) { | |
885 | ERR("Allocating socket"); | |
886 | goto error; | |
887 | } | |
888 | ||
889 | ret = lttcomm_create_sock(sock); | |
890 | if (ret < 0) { | |
891 | goto error; | |
892 | } | |
893 | DBG("Listening on sock %d", sock->fd); | |
894 | ||
895 | ret = sock->ops->bind(sock); | |
896 | if (ret < 0) { | |
2288467f | 897 | PERROR("Failed to bind socket"); |
b8aa1682 JD |
898 | goto error; |
899 | } | |
900 | ||
901 | ret = sock->ops->listen(sock, -1); | |
902 | if (ret < 0) { | |
903 | goto error; | |
904 | ||
905 | } | |
906 | ||
907 | return sock; | |
908 | ||
909 | error: | |
910 | if (sock) { | |
911 | lttcomm_destroy_sock(sock); | |
912 | } | |
913 | return NULL; | |
914 | } | |
915 | ||
916 | /* | |
917 | * This thread manages the listening for new connections on the network | |
918 | */ | |
7591bab1 | 919 | static void *relay_thread_listener(void *data) |
b8aa1682 | 920 | { |
095a4ae5 | 921 | int i, ret, pollfd, err = -1; |
b8aa1682 JD |
922 | uint32_t revents, nb_fd; |
923 | struct lttng_poll_event events; | |
924 | struct lttcomm_sock *control_sock, *data_sock; | |
925 | ||
b8aa1682 JD |
926 | DBG("[thread] Relay listener started"); |
927 | ||
55706a7d MD |
928 | health_register(health_relayd, HEALTH_RELAYD_TYPE_LISTENER); |
929 | ||
f385ae0a MD |
930 | health_code_update(); |
931 | ||
7591bab1 | 932 | control_sock = relay_socket_create(control_uri); |
b8aa1682 | 933 | if (!control_sock) { |
095a4ae5 | 934 | goto error_sock_control; |
b8aa1682 JD |
935 | } |
936 | ||
7591bab1 | 937 | data_sock = relay_socket_create(data_uri); |
b8aa1682 | 938 | if (!data_sock) { |
095a4ae5 | 939 | goto error_sock_relay; |
b8aa1682 JD |
940 | } |
941 | ||
942 | /* | |
7591bab1 MD |
943 | * Pass 3 as size here for the thread quit pipe, control and |
944 | * data socket. | |
b8aa1682 JD |
945 | */ |
946 | ret = create_thread_poll_set(&events, 3); | |
947 | if (ret < 0) { | |
948 | goto error_create_poll; | |
949 | } | |
950 | ||
951 | /* Add the control socket */ | |
952 | ret = lttng_poll_add(&events, control_sock->fd, LPOLLIN | LPOLLRDHUP); | |
953 | if (ret < 0) { | |
954 | goto error_poll_add; | |
955 | } | |
956 | ||
957 | /* Add the data socket */ | |
958 | ret = lttng_poll_add(&events, data_sock->fd, LPOLLIN | LPOLLRDHUP); | |
959 | if (ret < 0) { | |
960 | goto error_poll_add; | |
961 | } | |
962 | ||
3fd27398 MD |
963 | lttng_relay_notify_ready(); |
964 | ||
9b5e0863 MD |
965 | if (testpoint(relayd_thread_listener)) { |
966 | goto error_testpoint; | |
967 | } | |
968 | ||
b8aa1682 | 969 | while (1) { |
f385ae0a MD |
970 | health_code_update(); |
971 | ||
b8aa1682 JD |
972 | DBG("Listener accepting connections"); |
973 | ||
b8aa1682 | 974 | restart: |
f385ae0a | 975 | health_poll_entry(); |
b8aa1682 | 976 | ret = lttng_poll_wait(&events, -1); |
f385ae0a | 977 | health_poll_exit(); |
b8aa1682 JD |
978 | if (ret < 0) { |
979 | /* | |
980 | * Restart interrupted system call. | |
981 | */ | |
982 | if (errno == EINTR) { | |
983 | goto restart; | |
984 | } | |
985 | goto error; | |
986 | } | |
987 | ||
0d9c5d77 DG |
988 | nb_fd = ret; |
989 | ||
b8aa1682 JD |
990 | DBG("Relay new connection received"); |
991 | for (i = 0; i < nb_fd; i++) { | |
f385ae0a MD |
992 | health_code_update(); |
993 | ||
b8aa1682 JD |
994 | /* Fetch once the poll data */ |
995 | revents = LTTNG_POLL_GETEV(&events, i); | |
996 | pollfd = LTTNG_POLL_GETFD(&events, i); | |
997 | ||
998 | /* Thread quit pipe has been closed. Killing thread. */ | |
999 | ret = check_thread_quit_pipe(pollfd, revents); | |
1000 | if (ret) { | |
095a4ae5 MD |
1001 | err = 0; |
1002 | goto exit; | |
b8aa1682 JD |
1003 | } |
1004 | ||
03e43155 | 1005 | if (revents & LPOLLIN) { |
4b7f17b2 | 1006 | /* |
7591bab1 MD |
1007 | * A new connection is requested, therefore a |
1008 | * sessiond/consumerd connection is allocated in | |
1009 | * this thread, enqueued to a global queue and | |
1010 | * dequeued (and freed) in the worker thread. | |
4b7f17b2 | 1011 | */ |
58eb9381 DG |
1012 | int val = 1; |
1013 | struct relay_connection *new_conn; | |
4b7f17b2 | 1014 | struct lttcomm_sock *newsock; |
7591bab1 | 1015 | enum connection_type type; |
b8aa1682 JD |
1016 | |
1017 | if (pollfd == data_sock->fd) { | |
7591bab1 | 1018 | type = RELAY_DATA; |
b8aa1682 | 1019 | newsock = data_sock->ops->accept(data_sock); |
58eb9381 DG |
1020 | DBG("Relay data connection accepted, socket %d", |
1021 | newsock->fd); | |
4b7f17b2 MD |
1022 | } else { |
1023 | assert(pollfd == control_sock->fd); | |
7591bab1 | 1024 | type = RELAY_CONTROL; |
b8aa1682 | 1025 | newsock = control_sock->ops->accept(control_sock); |
58eb9381 DG |
1026 | DBG("Relay control connection accepted, socket %d", |
1027 | newsock->fd); | |
b8aa1682 | 1028 | } |
58eb9381 DG |
1029 | if (!newsock) { |
1030 | PERROR("accepting sock"); | |
58eb9381 DG |
1031 | goto error; |
1032 | } | |
1033 | ||
1034 | ret = setsockopt(newsock->fd, SOL_SOCKET, SO_REUSEADDR, &val, | |
1035 | sizeof(val)); | |
b8aa1682 JD |
1036 | if (ret < 0) { |
1037 | PERROR("setsockopt inet"); | |
4b7f17b2 | 1038 | lttcomm_destroy_sock(newsock); |
b8aa1682 JD |
1039 | goto error; |
1040 | } | |
f056029c JR |
1041 | |
1042 | ret = socket_apply_keep_alive_config(newsock->fd); | |
1043 | if (ret < 0) { | |
1044 | ERR("Failed to apply TCP keep-alive configuration on socket (%i)", | |
1045 | newsock->fd); | |
1046 | lttcomm_destroy_sock(newsock); | |
1047 | goto error; | |
1048 | } | |
1049 | ||
7591bab1 MD |
1050 | new_conn = connection_create(newsock, type); |
1051 | if (!new_conn) { | |
1052 | lttcomm_destroy_sock(newsock); | |
1053 | goto error; | |
1054 | } | |
58eb9381 DG |
1055 | |
1056 | /* Enqueue request for the dispatcher thread. */ | |
8bdee6e2 SM |
1057 | cds_wfcq_enqueue(&relay_conn_queue.head, &relay_conn_queue.tail, |
1058 | &new_conn->qnode); | |
b8aa1682 JD |
1059 | |
1060 | /* | |
7591bab1 MD |
1061 | * Wake the dispatch queue futex. |
1062 | * Implicit memory barrier with the | |
1063 | * exchange in cds_wfcq_enqueue. | |
b8aa1682 | 1064 | */ |
58eb9381 | 1065 | futex_nto1_wake(&relay_conn_queue.futex); |
03e43155 MD |
1066 | } else if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) { |
1067 | ERR("socket poll error"); | |
1068 | goto error; | |
1069 | } else { | |
1070 | ERR("Unexpected poll events %u for sock %d", revents, pollfd); | |
1071 | goto error; | |
b8aa1682 JD |
1072 | } |
1073 | } | |
1074 | } | |
1075 | ||
095a4ae5 | 1076 | exit: |
b8aa1682 JD |
1077 | error: |
1078 | error_poll_add: | |
9b5e0863 | 1079 | error_testpoint: |
b8aa1682 JD |
1080 | lttng_poll_clean(&events); |
1081 | error_create_poll: | |
095a4ae5 MD |
1082 | if (data_sock->fd >= 0) { |
1083 | ret = data_sock->ops->close(data_sock); | |
b8aa1682 JD |
1084 | if (ret) { |
1085 | PERROR("close"); | |
1086 | } | |
b8aa1682 | 1087 | } |
095a4ae5 MD |
1088 | lttcomm_destroy_sock(data_sock); |
1089 | error_sock_relay: | |
1090 | if (control_sock->fd >= 0) { | |
1091 | ret = control_sock->ops->close(control_sock); | |
b8aa1682 JD |
1092 | if (ret) { |
1093 | PERROR("close"); | |
1094 | } | |
b8aa1682 | 1095 | } |
095a4ae5 MD |
1096 | lttcomm_destroy_sock(control_sock); |
1097 | error_sock_control: | |
1098 | if (err) { | |
f385ae0a MD |
1099 | health_error(); |
1100 | ERR("Health error occurred in %s", __func__); | |
095a4ae5 | 1101 | } |
55706a7d | 1102 | health_unregister(health_relayd); |
b8aa1682 | 1103 | DBG("Relay listener thread cleanup complete"); |
b4aacfdc | 1104 | lttng_relay_stop_threads(); |
b8aa1682 JD |
1105 | return NULL; |
1106 | } | |
1107 | ||
1108 | /* | |
1109 | * This thread manages the dispatching of the requests to worker threads | |
1110 | */ | |
7591bab1 | 1111 | static void *relay_thread_dispatcher(void *data) |
b8aa1682 | 1112 | { |
6cd525e8 MD |
1113 | int err = -1; |
1114 | ssize_t ret; | |
8bdee6e2 | 1115 | struct cds_wfcq_node *node; |
58eb9381 | 1116 | struct relay_connection *new_conn = NULL; |
b8aa1682 JD |
1117 | |
1118 | DBG("[thread] Relay dispatcher started"); | |
1119 | ||
55706a7d MD |
1120 | health_register(health_relayd, HEALTH_RELAYD_TYPE_DISPATCHER); |
1121 | ||
9b5e0863 MD |
1122 | if (testpoint(relayd_thread_dispatcher)) { |
1123 | goto error_testpoint; | |
1124 | } | |
1125 | ||
f385ae0a MD |
1126 | health_code_update(); |
1127 | ||
0ed3b1a8 | 1128 | for (;;) { |
f385ae0a MD |
1129 | health_code_update(); |
1130 | ||
b8aa1682 | 1131 | /* Atomically prepare the queue futex */ |
58eb9381 | 1132 | futex_nto1_prepare(&relay_conn_queue.futex); |
b8aa1682 | 1133 | |
0ed3b1a8 MD |
1134 | if (CMM_LOAD_SHARED(dispatch_thread_exit)) { |
1135 | break; | |
1136 | } | |
1137 | ||
b8aa1682 | 1138 | do { |
f385ae0a MD |
1139 | health_code_update(); |
1140 | ||
b8aa1682 | 1141 | /* Dequeue commands */ |
8bdee6e2 SM |
1142 | node = cds_wfcq_dequeue_blocking(&relay_conn_queue.head, |
1143 | &relay_conn_queue.tail); | |
b8aa1682 JD |
1144 | if (node == NULL) { |
1145 | DBG("Woken up but nothing in the relay command queue"); | |
1146 | /* Continue thread execution */ | |
1147 | break; | |
1148 | } | |
58eb9381 | 1149 | new_conn = caa_container_of(node, struct relay_connection, qnode); |
b8aa1682 | 1150 | |
58eb9381 | 1151 | DBG("Dispatching request waiting on sock %d", new_conn->sock->fd); |
b8aa1682 JD |
1152 | |
1153 | /* | |
7591bab1 MD |
1154 | * Inform worker thread of the new request. This |
1155 | * call is blocking so we can be assured that | |
1156 | * the data will be read at some point in time | |
1157 | * or wait to the end of the world :) | |
b8aa1682 | 1158 | */ |
58eb9381 DG |
1159 | ret = lttng_write(relay_conn_pipe[1], &new_conn, sizeof(new_conn)); |
1160 | if (ret < 0) { | |
1161 | PERROR("write connection pipe"); | |
7591bab1 | 1162 | connection_put(new_conn); |
b8aa1682 JD |
1163 | goto error; |
1164 | } | |
1165 | } while (node != NULL); | |
1166 | ||
1167 | /* Futex wait on queue. Blocking call on futex() */ | |
f385ae0a | 1168 | health_poll_entry(); |
58eb9381 | 1169 | futex_nto1_wait(&relay_conn_queue.futex); |
f385ae0a | 1170 | health_poll_exit(); |
b8aa1682 JD |
1171 | } |
1172 | ||
f385ae0a MD |
1173 | /* Normal exit, no error */ |
1174 | err = 0; | |
1175 | ||
b8aa1682 | 1176 | error: |
9b5e0863 | 1177 | error_testpoint: |
f385ae0a MD |
1178 | if (err) { |
1179 | health_error(); | |
1180 | ERR("Health error occurred in %s", __func__); | |
1181 | } | |
55706a7d | 1182 | health_unregister(health_relayd); |
b8aa1682 | 1183 | DBG("Dispatch thread dying"); |
b4aacfdc | 1184 | lttng_relay_stop_threads(); |
b8aa1682 JD |
1185 | return NULL; |
1186 | } | |
1187 | ||
298a25ca JG |
1188 | static bool session_streams_have_index(const struct relay_session *session) |
1189 | { | |
1190 | return session->minor >= 4 && !session->snapshot; | |
1191 | } | |
1192 | ||
c5b6f4f0 DG |
1193 | /* |
1194 | * Handle the RELAYD_CREATE_SESSION command. | |
1195 | * | |
1196 | * On success, send back the session id or else return a negative value. | |
1197 | */ | |
5312a3ed JG |
1198 | static int relay_create_session(const struct lttcomm_relayd_hdr *recv_hdr, |
1199 | struct relay_connection *conn, | |
1200 | const struct lttng_buffer_view *payload) | |
c5b6f4f0 | 1201 | { |
5312a3ed JG |
1202 | int ret = 0; |
1203 | ssize_t send_ret; | |
4c6885d2 | 1204 | struct relay_session *session = NULL; |
ecd1a12f | 1205 | struct lttcomm_relayd_create_session_reply_2_11 reply = {}; |
1e791a74 JG |
1206 | char session_name[LTTNG_NAME_MAX] = {}; |
1207 | char hostname[LTTNG_HOST_NAME_MAX] = {}; | |
7591bab1 MD |
1208 | uint32_t live_timer = 0; |
1209 | bool snapshot = false; | |
46ef2188 | 1210 | bool session_name_contains_creation_timestamp = false; |
23c8ff50 | 1211 | /* Left nil for peers < 2.11. */ |
6fa5fe7c | 1212 | char base_path[LTTNG_PATH_MAX] = {}; |
23c8ff50 | 1213 | lttng_uuid sessiond_uuid = {}; |
1e791a74 JG |
1214 | LTTNG_OPTIONAL(uint64_t) id_sessiond = {}; |
1215 | LTTNG_OPTIONAL(uint64_t) current_chunk_id = {}; | |
db1da059 | 1216 | LTTNG_OPTIONAL(time_t) creation_time = {}; |
ecd1a12f MD |
1217 | struct lttng_dynamic_buffer reply_payload; |
1218 | ||
1219 | lttng_dynamic_buffer_init(&reply_payload); | |
c5b6f4f0 | 1220 | |
f86f6389 JR |
1221 | if (conn->minor < 4) { |
1222 | /* From 2.1 to 2.3 */ | |
1223 | ret = 0; | |
1224 | } else if (conn->minor >= 4 && conn->minor < 11) { | |
1225 | /* From 2.4 to 2.10 */ | |
5312a3ed | 1226 | ret = cmd_create_session_2_4(payload, session_name, |
7591bab1 | 1227 | hostname, &live_timer, &snapshot); |
f86f6389 | 1228 | } else { |
84fa4db5 | 1229 | bool has_current_chunk; |
db1da059 JG |
1230 | uint64_t current_chunk_id_value; |
1231 | time_t creation_time_value; | |
1232 | uint64_t id_sessiond_value; | |
84fa4db5 | 1233 | |
f86f6389 | 1234 | /* From 2.11 to ... */ |
db1da059 | 1235 | ret = cmd_create_session_2_11(payload, session_name, hostname, |
6fa5fe7c | 1236 | base_path, &live_timer, &snapshot, &id_sessiond_value, |
db1da059 | 1237 | sessiond_uuid, &has_current_chunk, |
46ef2188 MD |
1238 | ¤t_chunk_id_value, &creation_time_value, |
1239 | &session_name_contains_creation_timestamp); | |
23c8ff50 JG |
1240 | if (lttng_uuid_is_nil(sessiond_uuid)) { |
1241 | /* The nil UUID is reserved for pre-2.11 clients. */ | |
1242 | ERR("Illegal nil UUID announced by peer in create session command"); | |
1243 | ret = -1; | |
1244 | goto send_reply; | |
1245 | } | |
db1da059 JG |
1246 | LTTNG_OPTIONAL_SET(&id_sessiond, id_sessiond_value); |
1247 | LTTNG_OPTIONAL_SET(&creation_time, creation_time_value); | |
1248 | if (has_current_chunk) { | |
1249 | LTTNG_OPTIONAL_SET(¤t_chunk_id, | |
1250 | current_chunk_id_value); | |
1251 | } | |
7591bab1 | 1252 | } |
f86f6389 | 1253 | |
7591bab1 MD |
1254 | if (ret < 0) { |
1255 | goto send_reply; | |
d3e2ba59 JD |
1256 | } |
1257 | ||
6fa5fe7c | 1258 | session = session_create(session_name, hostname, base_path, live_timer, |
1e791a74 JG |
1259 | snapshot, sessiond_uuid, |
1260 | id_sessiond.is_set ? &id_sessiond.value : NULL, | |
1261 | current_chunk_id.is_set ? ¤t_chunk_id.value : NULL, | |
db1da059 | 1262 | creation_time.is_set ? &creation_time.value : NULL, |
46ef2188 MD |
1263 | conn->major, conn->minor, |
1264 | session_name_contains_creation_timestamp); | |
7591bab1 MD |
1265 | if (!session) { |
1266 | ret = -1; | |
1267 | goto send_reply; | |
1268 | } | |
1269 | assert(!conn->session); | |
1270 | conn->session = session; | |
c5b6f4f0 DG |
1271 | DBG("Created session %" PRIu64, session->id); |
1272 | ||
ecd1a12f | 1273 | reply.generic.session_id = htobe64(session->id); |
7591bab1 MD |
1274 | |
1275 | send_reply: | |
c5b6f4f0 | 1276 | if (ret < 0) { |
ecd1a12f | 1277 | reply.generic.ret_code = htobe32(LTTNG_ERR_FATAL); |
c5b6f4f0 | 1278 | } else { |
ecd1a12f | 1279 | reply.generic.ret_code = htobe32(LTTNG_OK); |
c5b6f4f0 DG |
1280 | } |
1281 | ||
ecd1a12f MD |
1282 | if (conn->minor < 11) { |
1283 | /* From 2.1 to 2.10 */ | |
1284 | ret = lttng_dynamic_buffer_append(&reply_payload, | |
1285 | &reply.generic, sizeof(reply.generic)); | |
1286 | if (ret) { | |
1287 | ERR("Failed to append \"create session\" command reply header to payload buffer"); | |
1288 | ret = -1; | |
1289 | goto end; | |
1290 | } | |
1291 | } else { | |
1292 | const uint32_t output_path_length = | |
8d382dd4 | 1293 | session ? strlen(session->output_path) + 1 : 0; |
ecd1a12f MD |
1294 | |
1295 | reply.output_path_length = htobe32(output_path_length); | |
8d382dd4 JG |
1296 | ret = lttng_dynamic_buffer_append( |
1297 | &reply_payload, &reply, sizeof(reply)); | |
ecd1a12f MD |
1298 | if (ret) { |
1299 | ERR("Failed to append \"create session\" command reply header to payload buffer"); | |
1300 | goto end; | |
1301 | } | |
1302 | ||
8d382dd4 JG |
1303 | if (output_path_length) { |
1304 | ret = lttng_dynamic_buffer_append(&reply_payload, | |
1305 | session->output_path, | |
1306 | output_path_length); | |
1307 | if (ret) { | |
1308 | ERR("Failed to append \"create session\" command reply path to payload buffer"); | |
1309 | goto end; | |
1310 | } | |
ecd1a12f MD |
1311 | } |
1312 | } | |
1313 | ||
1314 | send_ret = conn->sock->ops->sendmsg(conn->sock, reply_payload.data, | |
1315 | reply_payload.size, 0); | |
1316 | if (send_ret < (ssize_t) reply_payload.size) { | |
1317 | ERR("Failed to send \"create session\" command reply of %zu bytes (ret = %zd)", | |
1318 | reply_payload.size, send_ret); | |
5312a3ed | 1319 | ret = -1; |
c5b6f4f0 | 1320 | } |
ecd1a12f | 1321 | end: |
4c6885d2 JG |
1322 | if (ret < 0 && session) { |
1323 | session_put(session); | |
1324 | } | |
ecd1a12f | 1325 | lttng_dynamic_buffer_reset(&reply_payload); |
c5b6f4f0 DG |
1326 | return ret; |
1327 | } | |
1328 | ||
a4baae1b JD |
1329 | /* |
1330 | * When we have received all the streams and the metadata for a channel, | |
1331 | * we make them visible to the viewer threads. | |
1332 | */ | |
7591bab1 | 1333 | static void publish_connection_local_streams(struct relay_connection *conn) |
a4baae1b | 1334 | { |
7591bab1 MD |
1335 | struct relay_stream *stream; |
1336 | struct relay_session *session = conn->session; | |
a4baae1b | 1337 | |
7591bab1 MD |
1338 | /* |
1339 | * We publish all streams belonging to a session atomically wrt | |
1340 | * session lock. | |
1341 | */ | |
1342 | pthread_mutex_lock(&session->lock); | |
1343 | rcu_read_lock(); | |
1344 | cds_list_for_each_entry_rcu(stream, &session->recv_list, | |
1345 | recv_node) { | |
1346 | stream_publish(stream); | |
a4baae1b | 1347 | } |
7591bab1 | 1348 | rcu_read_unlock(); |
a4baae1b | 1349 | |
7591bab1 MD |
1350 | /* |
1351 | * Inform the viewer that there are new streams in the session. | |
1352 | */ | |
1353 | if (session->viewer_attached) { | |
1354 | uatomic_set(&session->new_streams, 1); | |
1355 | } | |
1356 | pthread_mutex_unlock(&session->lock); | |
a4baae1b JD |
1357 | } |
1358 | ||
348a81dc JG |
1359 | static int conform_channel_path(char *channel_path) |
1360 | { | |
1361 | int ret = 0; | |
1362 | ||
1363 | if (strstr("../", channel_path)) { | |
1364 | ERR("Refusing channel path as it walks up the path hierarchy: \"%s\"", | |
1365 | channel_path); | |
1366 | ret = -1; | |
1367 | goto end; | |
1368 | } | |
1369 | ||
1370 | if (*channel_path == '/') { | |
1371 | const size_t len = strlen(channel_path); | |
1372 | ||
1373 | /* | |
1374 | * Channel paths from peers prior to 2.11 are expressed as an | |
1375 | * absolute path that is, in reality, relative to the relay | |
1376 | * daemon's output directory. Remove the leading slash so it | |
1377 | * is correctly interpreted as a relative path later on. | |
1378 | * | |
1379 | * len (and not len - 1) is used to copy the trailing NULL. | |
1380 | */ | |
1381 | bcopy(channel_path + 1, channel_path, len); | |
1382 | } | |
1383 | end: | |
1384 | return ret; | |
1385 | } | |
1386 | ||
b8aa1682 JD |
1387 | /* |
1388 | * relay_add_stream: allocate a new stream for a session | |
1389 | */ | |
5312a3ed JG |
1390 | static int relay_add_stream(const struct lttcomm_relayd_hdr *recv_hdr, |
1391 | struct relay_connection *conn, | |
1392 | const struct lttng_buffer_view *payload) | |
b8aa1682 | 1393 | { |
7591bab1 MD |
1394 | int ret; |
1395 | ssize_t send_ret; | |
58eb9381 | 1396 | struct relay_session *session = conn->session; |
b8aa1682 JD |
1397 | struct relay_stream *stream = NULL; |
1398 | struct lttcomm_relayd_status_stream reply; | |
4030a636 | 1399 | struct ctf_trace *trace = NULL; |
7591bab1 MD |
1400 | uint64_t stream_handle = -1ULL; |
1401 | char *path_name = NULL, *channel_name = NULL; | |
1402 | uint64_t tracefile_size = 0, tracefile_count = 0; | |
348a81dc | 1403 | LTTNG_OPTIONAL(uint64_t) stream_chunk_id = {}; |
b8aa1682 | 1404 | |
5312a3ed | 1405 | if (!session || !conn->version_check_done) { |
b8aa1682 JD |
1406 | ERR("Trying to add a stream before version check"); |
1407 | ret = -1; | |
1408 | goto end_no_session; | |
1409 | } | |
1410 | ||
2f21a469 JR |
1411 | if (session->minor == 1) { |
1412 | /* For 2.1 */ | |
5312a3ed | 1413 | ret = cmd_recv_stream_2_1(payload, &path_name, |
7591bab1 | 1414 | &channel_name); |
2f21a469 JR |
1415 | } else if (session->minor > 1 && session->minor < 11) { |
1416 | /* From 2.2 to 2.10 */ | |
5312a3ed | 1417 | ret = cmd_recv_stream_2_2(payload, &path_name, |
7591bab1 | 1418 | &channel_name, &tracefile_size, &tracefile_count); |
2f21a469 JR |
1419 | } else { |
1420 | /* From 2.11 to ... */ | |
1421 | ret = cmd_recv_stream_2_11(payload, &path_name, | |
0b50e4b3 JG |
1422 | &channel_name, &tracefile_size, &tracefile_count, |
1423 | &stream_chunk_id.value); | |
1424 | stream_chunk_id.is_set = true; | |
0f907de1 | 1425 | } |
2f21a469 | 1426 | |
0f907de1 | 1427 | if (ret < 0) { |
7591bab1 | 1428 | goto send_reply; |
b8aa1682 JD |
1429 | } |
1430 | ||
348a81dc JG |
1431 | if (conform_channel_path(path_name)) { |
1432 | goto send_reply; | |
1433 | } | |
1434 | ||
2a635488 JR |
1435 | /* |
1436 | * Backward compatibility for --group-output-by-session. | |
1437 | * Prior to lttng 2.11, the complete path is passed by the stream. | |
1438 | * Starting at 2.11, lttng-relayd uses chunk. When dealing with producer | |
1439 | * >=2.11 the chunk is responsible for the output path. When dealing | |
1440 | * with producer < 2.11 the chunk output_path is the root output path | |
1441 | * and the stream carries the complete path (path_name). | |
1442 | * To support --group-output-by-session with older producer (<2.11), we | |
1443 | * need to craft the path based on the stream path. | |
1444 | */ | |
1445 | if (opt_group_output_by == RELAYD_GROUP_OUTPUT_BY_SESSION) { | |
1446 | if (conn->minor < 4) { | |
1447 | /* | |
1448 | * From 2.1 to 2.3, the session_name is not passed on | |
1449 | * the RELAYD_CREATE_SESSION command. The session name | |
1450 | * is necessary to detect the presence of a base_path | |
1451 | * inside the stream path. Without it we cannot perform | |
1452 | * a valid group-output-by-session transformation. | |
1453 | */ | |
1454 | WARN("Unable to perform a --group-by-session transformation for session %" PRIu64 | |
1455 | " for stream with path \"%s\" as it is produced by a peer using a protocol older than v2.4", | |
1456 | session->id, path_name); | |
1457 | } else if (conn->minor >= 4 && conn->minor < 11) { | |
1458 | char *group_by_session_path_name; | |
1459 | ||
1460 | assert(session->session_name[0] != '\0'); | |
1461 | ||
1462 | group_by_session_path_name = | |
1463 | backward_compat_group_by_session( | |
1464 | path_name, | |
1465 | session->session_name); | |
1466 | if (!group_by_session_path_name) { | |
1467 | ERR("Failed to apply group by session to stream of session %" PRIu64, | |
1468 | session->id); | |
1469 | goto send_reply; | |
1470 | } | |
1471 | ||
1472 | DBG("Transformed session path from \"%s\" to \"%s\" to honor per-session name grouping", | |
1473 | path_name, group_by_session_path_name); | |
1474 | ||
1475 | free(path_name); | |
1476 | path_name = group_by_session_path_name; | |
1477 | } | |
1478 | } | |
1479 | ||
7591bab1 | 1480 | trace = ctf_trace_get_by_path_or_create(session, path_name); |
2a174661 | 1481 | if (!trace) { |
7591bab1 | 1482 | goto send_reply; |
2a174661 | 1483 | } |
2a174661 | 1484 | |
2a635488 | 1485 | /* This stream here has one reference on the trace. */ |
7591bab1 MD |
1486 | pthread_mutex_lock(&last_relay_stream_id_lock); |
1487 | stream_handle = ++last_relay_stream_id; | |
1488 | pthread_mutex_unlock(&last_relay_stream_id_lock); | |
d3e2ba59 | 1489 | |
7591bab1 MD |
1490 | /* We pass ownership of path_name and channel_name. */ |
1491 | stream = stream_create(trace, stream_handle, path_name, | |
348a81dc | 1492 | channel_name, tracefile_size, tracefile_count); |
7591bab1 MD |
1493 | path_name = NULL; |
1494 | channel_name = NULL; | |
a4baae1b | 1495 | |
2a174661 | 1496 | /* |
7591bab1 MD |
1497 | * Streams are the owners of their trace. Reference to trace is |
1498 | * kept within stream_create(). | |
2a174661 | 1499 | */ |
7591bab1 | 1500 | ctf_trace_put(trace); |
d3e2ba59 | 1501 | |
7591bab1 | 1502 | send_reply: |
53efb85a | 1503 | memset(&reply, 0, sizeof(reply)); |
7591bab1 MD |
1504 | reply.handle = htobe64(stream_handle); |
1505 | if (!stream) { | |
f73fabfd | 1506 | reply.ret_code = htobe32(LTTNG_ERR_UNK); |
b8aa1682 | 1507 | } else { |
f73fabfd | 1508 | reply.ret_code = htobe32(LTTNG_OK); |
b8aa1682 | 1509 | } |
5af40280 | 1510 | |
58eb9381 | 1511 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, |
b8aa1682 | 1512 | sizeof(struct lttcomm_relayd_status_stream), 0); |
5312a3ed JG |
1513 | if (send_ret < (ssize_t) sizeof(reply)) { |
1514 | ERR("Failed to send \"add stream\" command reply (ret = %zd)", | |
1515 | send_ret); | |
1516 | ret = -1; | |
b8aa1682 JD |
1517 | } |
1518 | ||
1519 | end_no_session: | |
7591bab1 MD |
1520 | free(path_name); |
1521 | free(channel_name); | |
0f907de1 | 1522 | return ret; |
b8aa1682 JD |
1523 | } |
1524 | ||
173af62f DG |
1525 | /* |
1526 | * relay_close_stream: close a specific stream | |
1527 | */ | |
5312a3ed JG |
1528 | static int relay_close_stream(const struct lttcomm_relayd_hdr *recv_hdr, |
1529 | struct relay_connection *conn, | |
1530 | const struct lttng_buffer_view *payload) | |
173af62f | 1531 | { |
5312a3ed JG |
1532 | int ret; |
1533 | ssize_t send_ret; | |
58eb9381 | 1534 | struct relay_session *session = conn->session; |
173af62f DG |
1535 | struct lttcomm_relayd_close_stream stream_info; |
1536 | struct lttcomm_relayd_generic_reply reply; | |
1537 | struct relay_stream *stream; | |
173af62f DG |
1538 | |
1539 | DBG("Close stream received"); | |
1540 | ||
5312a3ed | 1541 | if (!session || !conn->version_check_done) { |
173af62f DG |
1542 | ERR("Trying to close a stream before version check"); |
1543 | ret = -1; | |
1544 | goto end_no_session; | |
1545 | } | |
1546 | ||
5312a3ed JG |
1547 | if (payload->size < sizeof(stream_info)) { |
1548 | ERR("Unexpected payload size in \"relay_close_stream\": expected >= %zu bytes, got %zu bytes", | |
1549 | sizeof(stream_info), payload->size); | |
173af62f DG |
1550 | ret = -1; |
1551 | goto end_no_session; | |
1552 | } | |
5312a3ed JG |
1553 | memcpy(&stream_info, payload->data, sizeof(stream_info)); |
1554 | stream_info.stream_id = be64toh(stream_info.stream_id); | |
1555 | stream_info.last_net_seq_num = be64toh(stream_info.last_net_seq_num); | |
173af62f | 1556 | |
5312a3ed | 1557 | stream = stream_get_by_id(stream_info.stream_id); |
173af62f DG |
1558 | if (!stream) { |
1559 | ret = -1; | |
7591bab1 | 1560 | goto end; |
173af62f | 1561 | } |
77f7bd85 MD |
1562 | |
1563 | /* | |
1564 | * Set last_net_seq_num before the close flag. Required by data | |
1565 | * pending check. | |
1566 | */ | |
7591bab1 | 1567 | pthread_mutex_lock(&stream->lock); |
5312a3ed | 1568 | stream->last_net_seq_num = stream_info.last_net_seq_num; |
77f7bd85 MD |
1569 | pthread_mutex_unlock(&stream->lock); |
1570 | ||
bda7c7b9 JG |
1571 | /* |
1572 | * This is one of the conditions which may trigger a stream close | |
1573 | * with the others being: | |
1574 | * 1) A close command is received for a stream | |
1575 | * 2) The control connection owning the stream is closed | |
1576 | * 3) We have received all of the stream's data _after_ a close | |
1577 | * request. | |
1578 | */ | |
1579 | try_stream_close(stream); | |
7591bab1 MD |
1580 | if (stream->is_metadata) { |
1581 | struct relay_viewer_stream *vstream; | |
173af62f | 1582 | |
7591bab1 MD |
1583 | vstream = viewer_stream_get_by_id(stream->stream_handle); |
1584 | if (vstream) { | |
94f73d08 | 1585 | if (stream->no_new_metadata_notified) { |
7591bab1 MD |
1586 | /* |
1587 | * Since all the metadata has been sent to the | |
1588 | * viewer and that we have a request to close | |
1589 | * its stream, we can safely teardown the | |
1590 | * corresponding metadata viewer stream. | |
1591 | */ | |
1592 | viewer_stream_put(vstream); | |
1593 | } | |
1594 | /* Put local reference. */ | |
1595 | viewer_stream_put(vstream); | |
1596 | } | |
1597 | } | |
7591bab1 | 1598 | stream_put(stream); |
5312a3ed | 1599 | ret = 0; |
173af62f | 1600 | |
7591bab1 | 1601 | end: |
53efb85a | 1602 | memset(&reply, 0, sizeof(reply)); |
173af62f | 1603 | if (ret < 0) { |
f73fabfd | 1604 | reply.ret_code = htobe32(LTTNG_ERR_UNK); |
173af62f | 1605 | } else { |
f73fabfd | 1606 | reply.ret_code = htobe32(LTTNG_OK); |
173af62f | 1607 | } |
58eb9381 | 1608 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, |
173af62f | 1609 | sizeof(struct lttcomm_relayd_generic_reply), 0); |
5312a3ed JG |
1610 | if (send_ret < (ssize_t) sizeof(reply)) { |
1611 | ERR("Failed to send \"close stream\" command reply (ret = %zd)", | |
1612 | send_ret); | |
1613 | ret = -1; | |
173af62f DG |
1614 | } |
1615 | ||
1616 | end_no_session: | |
1617 | return ret; | |
1618 | } | |
1619 | ||
93ec662e JD |
1620 | /* |
1621 | * relay_reset_metadata: reset a metadata stream | |
1622 | */ | |
1623 | static | |
5312a3ed JG |
1624 | int relay_reset_metadata(const struct lttcomm_relayd_hdr *recv_hdr, |
1625 | struct relay_connection *conn, | |
1626 | const struct lttng_buffer_view *payload) | |
93ec662e | 1627 | { |
5312a3ed JG |
1628 | int ret; |
1629 | ssize_t send_ret; | |
93ec662e JD |
1630 | struct relay_session *session = conn->session; |
1631 | struct lttcomm_relayd_reset_metadata stream_info; | |
1632 | struct lttcomm_relayd_generic_reply reply; | |
1633 | struct relay_stream *stream; | |
1634 | ||
1635 | DBG("Reset metadata received"); | |
1636 | ||
5312a3ed | 1637 | if (!session || !conn->version_check_done) { |
93ec662e JD |
1638 | ERR("Trying to reset a metadata stream before version check"); |
1639 | ret = -1; | |
1640 | goto end_no_session; | |
1641 | } | |
1642 | ||
5312a3ed JG |
1643 | if (payload->size < sizeof(stream_info)) { |
1644 | ERR("Unexpected payload size in \"relay_reset_metadata\": expected >= %zu bytes, got %zu bytes", | |
1645 | sizeof(stream_info), payload->size); | |
93ec662e JD |
1646 | ret = -1; |
1647 | goto end_no_session; | |
1648 | } | |
5312a3ed JG |
1649 | memcpy(&stream_info, payload->data, sizeof(stream_info)); |
1650 | stream_info.stream_id = be64toh(stream_info.stream_id); | |
1651 | stream_info.version = be64toh(stream_info.version); | |
1652 | ||
1653 | DBG("Update metadata to version %" PRIu64, stream_info.version); | |
93ec662e JD |
1654 | |
1655 | /* Unsupported for live sessions for now. */ | |
1656 | if (session->live_timer != 0) { | |
1657 | ret = -1; | |
1658 | goto end; | |
1659 | } | |
1660 | ||
5312a3ed | 1661 | stream = stream_get_by_id(stream_info.stream_id); |
93ec662e JD |
1662 | if (!stream) { |
1663 | ret = -1; | |
1664 | goto end; | |
1665 | } | |
1666 | pthread_mutex_lock(&stream->lock); | |
1667 | if (!stream->is_metadata) { | |
1668 | ret = -1; | |
1669 | goto end_unlock; | |
1670 | } | |
1671 | ||
c35f9726 | 1672 | ret = stream_reset_file(stream); |
93ec662e | 1673 | if (ret < 0) { |
c35f9726 JG |
1674 | ERR("Failed to reset metadata stream %" PRIu64 |
1675 | ": stream_path = %s, channel = %s", | |
1676 | stream->stream_handle, stream->path_name, | |
1677 | stream->channel_name); | |
93ec662e JD |
1678 | goto end_unlock; |
1679 | } | |
93ec662e JD |
1680 | end_unlock: |
1681 | pthread_mutex_unlock(&stream->lock); | |
1682 | stream_put(stream); | |
1683 | ||
1684 | end: | |
1685 | memset(&reply, 0, sizeof(reply)); | |
1686 | if (ret < 0) { | |
1687 | reply.ret_code = htobe32(LTTNG_ERR_UNK); | |
1688 | } else { | |
1689 | reply.ret_code = htobe32(LTTNG_OK); | |
1690 | } | |
1691 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, | |
1692 | sizeof(struct lttcomm_relayd_generic_reply), 0); | |
5312a3ed JG |
1693 | if (send_ret < (ssize_t) sizeof(reply)) { |
1694 | ERR("Failed to send \"reset metadata\" command reply (ret = %zd)", | |
1695 | send_ret); | |
1696 | ret = -1; | |
93ec662e JD |
1697 | } |
1698 | ||
1699 | end_no_session: | |
1700 | return ret; | |
1701 | } | |
1702 | ||
b8aa1682 JD |
1703 | /* |
1704 | * relay_unknown_command: send -1 if received unknown command | |
1705 | */ | |
7591bab1 | 1706 | static void relay_unknown_command(struct relay_connection *conn) |
b8aa1682 JD |
1707 | { |
1708 | struct lttcomm_relayd_generic_reply reply; | |
5312a3ed | 1709 | ssize_t send_ret; |
b8aa1682 | 1710 | |
53efb85a | 1711 | memset(&reply, 0, sizeof(reply)); |
f73fabfd | 1712 | reply.ret_code = htobe32(LTTNG_ERR_UNK); |
5312a3ed JG |
1713 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
1714 | if (send_ret < sizeof(reply)) { | |
1715 | ERR("Failed to send \"unknown command\" command reply (ret = %zd)", send_ret); | |
b8aa1682 JD |
1716 | } |
1717 | } | |
1718 | ||
1719 | /* | |
1720 | * relay_start: send an acknowledgment to the client to tell if we are | |
1721 | * ready to receive data. We are ready if a session is established. | |
1722 | */ | |
5312a3ed JG |
1723 | static int relay_start(const struct lttcomm_relayd_hdr *recv_hdr, |
1724 | struct relay_connection *conn, | |
1725 | const struct lttng_buffer_view *payload) | |
b8aa1682 | 1726 | { |
5312a3ed JG |
1727 | int ret = 0; |
1728 | ssize_t send_ret; | |
b8aa1682 | 1729 | struct lttcomm_relayd_generic_reply reply; |
58eb9381 | 1730 | struct relay_session *session = conn->session; |
b8aa1682 JD |
1731 | |
1732 | if (!session) { | |
1733 | DBG("Trying to start the streaming without a session established"); | |
f73fabfd | 1734 | ret = htobe32(LTTNG_ERR_UNK); |
b8aa1682 JD |
1735 | } |
1736 | ||
53efb85a | 1737 | memset(&reply, 0, sizeof(reply)); |
5312a3ed JG |
1738 | reply.ret_code = htobe32(LTTNG_OK); |
1739 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, | |
1740 | sizeof(reply), 0); | |
1741 | if (send_ret < (ssize_t) sizeof(reply)) { | |
1742 | ERR("Failed to send \"relay_start\" command reply (ret = %zd)", | |
1743 | send_ret); | |
1744 | ret = -1; | |
b8aa1682 JD |
1745 | } |
1746 | ||
1747 | return ret; | |
1748 | } | |
1749 | ||
b8aa1682 | 1750 | /* |
7591bab1 | 1751 | * relay_recv_metadata: receive the metadata for the session. |
b8aa1682 | 1752 | */ |
5312a3ed JG |
1753 | static int relay_recv_metadata(const struct lttcomm_relayd_hdr *recv_hdr, |
1754 | struct relay_connection *conn, | |
1755 | const struct lttng_buffer_view *payload) | |
b8aa1682 | 1756 | { |
32d1569c | 1757 | int ret = 0; |
58eb9381 | 1758 | struct relay_session *session = conn->session; |
5312a3ed | 1759 | struct lttcomm_relayd_metadata_payload metadata_payload_header; |
b8aa1682 | 1760 | struct relay_stream *metadata_stream; |
5312a3ed | 1761 | uint64_t metadata_payload_size; |
c35f9726 | 1762 | struct lttng_buffer_view packet_view; |
b8aa1682 JD |
1763 | |
1764 | if (!session) { | |
1765 | ERR("Metadata sent before version check"); | |
1766 | ret = -1; | |
1767 | goto end; | |
1768 | } | |
1769 | ||
5312a3ed | 1770 | if (recv_hdr->data_size < sizeof(struct lttcomm_relayd_metadata_payload)) { |
f6416125 MD |
1771 | ERR("Incorrect data size"); |
1772 | ret = -1; | |
1773 | goto end; | |
1774 | } | |
5312a3ed JG |
1775 | metadata_payload_size = recv_hdr->data_size - |
1776 | sizeof(struct lttcomm_relayd_metadata_payload); | |
f6416125 | 1777 | |
5312a3ed JG |
1778 | memcpy(&metadata_payload_header, payload->data, |
1779 | sizeof(metadata_payload_header)); | |
1780 | metadata_payload_header.stream_id = be64toh( | |
1781 | metadata_payload_header.stream_id); | |
1782 | metadata_payload_header.padding_size = be32toh( | |
1783 | metadata_payload_header.padding_size); | |
9d1bbf21 | 1784 | |
5312a3ed | 1785 | metadata_stream = stream_get_by_id(metadata_payload_header.stream_id); |
b8aa1682 JD |
1786 | if (!metadata_stream) { |
1787 | ret = -1; | |
7591bab1 | 1788 | goto end; |
b8aa1682 JD |
1789 | } |
1790 | ||
c35f9726 JG |
1791 | packet_view = lttng_buffer_view_from_view(payload, |
1792 | sizeof(metadata_payload_header), metadata_payload_size); | |
1793 | if (!packet_view.data) { | |
1794 | ERR("Invalid metadata packet length announced by header"); | |
b8aa1682 | 1795 | ret = -1; |
7591bab1 | 1796 | goto end_put; |
b8aa1682 | 1797 | } |
1d4dfdef | 1798 | |
c35f9726 JG |
1799 | pthread_mutex_lock(&metadata_stream->lock); |
1800 | ret = stream_write(metadata_stream, &packet_view, | |
5312a3ed | 1801 | metadata_payload_header.padding_size); |
c35f9726 JG |
1802 | pthread_mutex_unlock(&metadata_stream->lock); |
1803 | if (ret){ | |
5312a3ed | 1804 | ret = -1; |
7591bab1 | 1805 | goto end_put; |
1d4dfdef | 1806 | } |
7591bab1 | 1807 | end_put: |
7591bab1 | 1808 | stream_put(metadata_stream); |
b8aa1682 JD |
1809 | end: |
1810 | return ret; | |
1811 | } | |
1812 | ||
1813 | /* | |
1814 | * relay_send_version: send relayd version number | |
1815 | */ | |
5312a3ed JG |
1816 | static int relay_send_version(const struct lttcomm_relayd_hdr *recv_hdr, |
1817 | struct relay_connection *conn, | |
1818 | const struct lttng_buffer_view *payload) | |
b8aa1682 | 1819 | { |
7f51dcba | 1820 | int ret; |
5312a3ed | 1821 | ssize_t send_ret; |
092b6259 | 1822 | struct lttcomm_relayd_version reply, msg; |
87cb6359 | 1823 | bool compatible = true; |
b8aa1682 | 1824 | |
5312a3ed | 1825 | conn->version_check_done = true; |
b8aa1682 | 1826 | |
092b6259 | 1827 | /* Get version from the other side. */ |
5312a3ed JG |
1828 | if (payload->size < sizeof(msg)) { |
1829 | ERR("Unexpected payload size in \"relay_send_version\": expected >= %zu bytes, got %zu bytes", | |
1830 | sizeof(msg), payload->size); | |
092b6259 | 1831 | ret = -1; |
092b6259 DG |
1832 | goto end; |
1833 | } | |
1834 | ||
5312a3ed JG |
1835 | memcpy(&msg, payload->data, sizeof(msg)); |
1836 | msg.major = be32toh(msg.major); | |
1837 | msg.minor = be32toh(msg.minor); | |
1838 | ||
53efb85a | 1839 | memset(&reply, 0, sizeof(reply)); |
d83a952c MD |
1840 | reply.major = RELAYD_VERSION_COMM_MAJOR; |
1841 | reply.minor = RELAYD_VERSION_COMM_MINOR; | |
d4519fa3 JD |
1842 | |
1843 | /* Major versions must be the same */ | |
5312a3ed | 1844 | if (reply.major != msg.major) { |
6151a90f | 1845 | DBG("Incompatible major versions (%u vs %u), deleting session", |
5312a3ed | 1846 | reply.major, msg.major); |
87cb6359 | 1847 | compatible = false; |
d4519fa3 JD |
1848 | } |
1849 | ||
58eb9381 | 1850 | conn->major = reply.major; |
0f907de1 | 1851 | /* We adapt to the lowest compatible version */ |
5312a3ed | 1852 | if (reply.minor <= msg.minor) { |
58eb9381 | 1853 | conn->minor = reply.minor; |
0f907de1 | 1854 | } else { |
5312a3ed | 1855 | conn->minor = msg.minor; |
0f907de1 JD |
1856 | } |
1857 | ||
6151a90f JD |
1858 | reply.major = htobe32(reply.major); |
1859 | reply.minor = htobe32(reply.minor); | |
5312a3ed JG |
1860 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, |
1861 | sizeof(reply), 0); | |
1862 | if (send_ret < (ssize_t) sizeof(reply)) { | |
1863 | ERR("Failed to send \"send version\" command reply (ret = %zd)", | |
1864 | send_ret); | |
1865 | ret = -1; | |
1866 | goto end; | |
1867 | } else { | |
1868 | ret = 0; | |
6151a90f JD |
1869 | } |
1870 | ||
87cb6359 JD |
1871 | if (!compatible) { |
1872 | ret = -1; | |
1873 | goto end; | |
1874 | } | |
1875 | ||
58eb9381 DG |
1876 | DBG("Version check done using protocol %u.%u", conn->major, |
1877 | conn->minor); | |
b8aa1682 JD |
1878 | |
1879 | end: | |
1880 | return ret; | |
1881 | } | |
1882 | ||
c8f59ee5 | 1883 | /* |
6d805429 | 1884 | * Check for data pending for a given stream id from the session daemon. |
c8f59ee5 | 1885 | */ |
5312a3ed JG |
1886 | static int relay_data_pending(const struct lttcomm_relayd_hdr *recv_hdr, |
1887 | struct relay_connection *conn, | |
1888 | const struct lttng_buffer_view *payload) | |
c8f59ee5 | 1889 | { |
58eb9381 | 1890 | struct relay_session *session = conn->session; |
6d805429 | 1891 | struct lttcomm_relayd_data_pending msg; |
c8f59ee5 DG |
1892 | struct lttcomm_relayd_generic_reply reply; |
1893 | struct relay_stream *stream; | |
5312a3ed | 1894 | ssize_t send_ret; |
c8f59ee5 | 1895 | int ret; |
298a25ca | 1896 | uint64_t stream_seq; |
c8f59ee5 | 1897 | |
6d805429 | 1898 | DBG("Data pending command received"); |
c8f59ee5 | 1899 | |
5312a3ed | 1900 | if (!session || !conn->version_check_done) { |
c8f59ee5 DG |
1901 | ERR("Trying to check for data before version check"); |
1902 | ret = -1; | |
1903 | goto end_no_session; | |
1904 | } | |
1905 | ||
5312a3ed JG |
1906 | if (payload->size < sizeof(msg)) { |
1907 | ERR("Unexpected payload size in \"relay_data_pending\": expected >= %zu bytes, got %zu bytes", | |
1908 | sizeof(msg), payload->size); | |
c8f59ee5 DG |
1909 | ret = -1; |
1910 | goto end_no_session; | |
1911 | } | |
5312a3ed JG |
1912 | memcpy(&msg, payload->data, sizeof(msg)); |
1913 | msg.stream_id = be64toh(msg.stream_id); | |
1914 | msg.last_net_seq_num = be64toh(msg.last_net_seq_num); | |
c8f59ee5 | 1915 | |
5312a3ed | 1916 | stream = stream_get_by_id(msg.stream_id); |
de91f48a | 1917 | if (stream == NULL) { |
c8f59ee5 | 1918 | ret = -1; |
7591bab1 | 1919 | goto end; |
c8f59ee5 DG |
1920 | } |
1921 | ||
7591bab1 MD |
1922 | pthread_mutex_lock(&stream->lock); |
1923 | ||
298a25ca JG |
1924 | if (session_streams_have_index(session)) { |
1925 | /* | |
1926 | * Ensure that both the index and stream data have been | |
1927 | * flushed up to the requested point. | |
1928 | */ | |
a8f9f353 | 1929 | stream_seq = min(stream->prev_data_seq, stream->prev_index_seq); |
298a25ca | 1930 | } else { |
a8f9f353 | 1931 | stream_seq = stream->prev_data_seq; |
298a25ca | 1932 | } |
a8f9f353 | 1933 | DBG("Data pending for stream id %" PRIu64 ": prev_data_seq %" PRIu64 |
298a25ca JG |
1934 | ", prev_index_seq %" PRIu64 |
1935 | ", and last_seq %" PRIu64, msg.stream_id, | |
a8f9f353 | 1936 | stream->prev_data_seq, stream->prev_index_seq, |
298a25ca | 1937 | msg.last_net_seq_num); |
c8f59ee5 | 1938 | |
33832e64 | 1939 | /* Avoid wrapping issue */ |
298a25ca | 1940 | if (((int64_t) (stream_seq - msg.last_net_seq_num)) >= 0) { |
6d805429 | 1941 | /* Data has in fact been written and is NOT pending */ |
c8f59ee5 | 1942 | ret = 0; |
6d805429 DG |
1943 | } else { |
1944 | /* Data still being streamed thus pending */ | |
1945 | ret = 1; | |
c8f59ee5 DG |
1946 | } |
1947 | ||
7591bab1 MD |
1948 | stream->data_pending_check_done = true; |
1949 | pthread_mutex_unlock(&stream->lock); | |
f7079f67 | 1950 | |
7591bab1 MD |
1951 | stream_put(stream); |
1952 | end: | |
c8f59ee5 | 1953 | |
53efb85a | 1954 | memset(&reply, 0, sizeof(reply)); |
c8f59ee5 | 1955 | reply.ret_code = htobe32(ret); |
5312a3ed JG |
1956 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
1957 | if (send_ret < (ssize_t) sizeof(reply)) { | |
1958 | ERR("Failed to send \"data pending\" command reply (ret = %zd)", | |
1959 | send_ret); | |
1960 | ret = -1; | |
c8f59ee5 DG |
1961 | } |
1962 | ||
1963 | end_no_session: | |
1964 | return ret; | |
1965 | } | |
1966 | ||
1967 | /* | |
1968 | * Wait for the control socket to reach a quiescent state. | |
1969 | * | |
7591bab1 MD |
1970 | * Note that for now, when receiving this command from the session |
1971 | * daemon, this means that every subsequent commands or data received on | |
1972 | * the control socket has been handled. So, this is why we simply return | |
1973 | * OK here. | |
c8f59ee5 | 1974 | */ |
5312a3ed JG |
1975 | static int relay_quiescent_control(const struct lttcomm_relayd_hdr *recv_hdr, |
1976 | struct relay_connection *conn, | |
1977 | const struct lttng_buffer_view *payload) | |
c8f59ee5 DG |
1978 | { |
1979 | int ret; | |
5312a3ed | 1980 | ssize_t send_ret; |
ad7051c0 | 1981 | struct relay_stream *stream; |
ad7051c0 | 1982 | struct lttcomm_relayd_quiescent_control msg; |
c8f59ee5 DG |
1983 | struct lttcomm_relayd_generic_reply reply; |
1984 | ||
1985 | DBG("Checking quiescent state on control socket"); | |
1986 | ||
5312a3ed | 1987 | if (!conn->session || !conn->version_check_done) { |
ad7051c0 DG |
1988 | ERR("Trying to check for data before version check"); |
1989 | ret = -1; | |
1990 | goto end_no_session; | |
1991 | } | |
1992 | ||
5312a3ed JG |
1993 | if (payload->size < sizeof(msg)) { |
1994 | ERR("Unexpected payload size in \"relay_quiescent_control\": expected >= %zu bytes, got %zu bytes", | |
1995 | sizeof(msg), payload->size); | |
ad7051c0 DG |
1996 | ret = -1; |
1997 | goto end_no_session; | |
1998 | } | |
5312a3ed JG |
1999 | memcpy(&msg, payload->data, sizeof(msg)); |
2000 | msg.stream_id = be64toh(msg.stream_id); | |
ad7051c0 | 2001 | |
5312a3ed | 2002 | stream = stream_get_by_id(msg.stream_id); |
7591bab1 MD |
2003 | if (!stream) { |
2004 | goto reply; | |
2005 | } | |
2006 | pthread_mutex_lock(&stream->lock); | |
2007 | stream->data_pending_check_done = true; | |
2008 | pthread_mutex_unlock(&stream->lock); | |
5312a3ed JG |
2009 | |
2010 | DBG("Relay quiescent control pending flag set to %" PRIu64, msg.stream_id); | |
7591bab1 MD |
2011 | stream_put(stream); |
2012 | reply: | |
53efb85a | 2013 | memset(&reply, 0, sizeof(reply)); |
c8f59ee5 | 2014 | reply.ret_code = htobe32(LTTNG_OK); |
5312a3ed JG |
2015 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
2016 | if (send_ret < (ssize_t) sizeof(reply)) { | |
2017 | ERR("Failed to send \"quiescent control\" command reply (ret = %zd)", | |
2018 | send_ret); | |
2019 | ret = -1; | |
2020 | } else { | |
2021 | ret = 0; | |
c8f59ee5 DG |
2022 | } |
2023 | ||
ad7051c0 | 2024 | end_no_session: |
c8f59ee5 DG |
2025 | return ret; |
2026 | } | |
2027 | ||
f7079f67 | 2028 | /* |
7591bab1 MD |
2029 | * Initialize a data pending command. This means that a consumer is about |
2030 | * to ask for data pending for each stream it holds. Simply iterate over | |
2031 | * all streams of a session and set the data_pending_check_done flag. | |
f7079f67 DG |
2032 | * |
2033 | * This command returns to the client a LTTNG_OK code. | |
2034 | */ | |
5312a3ed JG |
2035 | static int relay_begin_data_pending(const struct lttcomm_relayd_hdr *recv_hdr, |
2036 | struct relay_connection *conn, | |
2037 | const struct lttng_buffer_view *payload) | |
f7079f67 DG |
2038 | { |
2039 | int ret; | |
5312a3ed | 2040 | ssize_t send_ret; |
f7079f67 DG |
2041 | struct lttng_ht_iter iter; |
2042 | struct lttcomm_relayd_begin_data_pending msg; | |
2043 | struct lttcomm_relayd_generic_reply reply; | |
2044 | struct relay_stream *stream; | |
f7079f67 DG |
2045 | |
2046 | assert(recv_hdr); | |
58eb9381 | 2047 | assert(conn); |
f7079f67 DG |
2048 | |
2049 | DBG("Init streams for data pending"); | |
2050 | ||
5312a3ed | 2051 | if (!conn->session || !conn->version_check_done) { |
f7079f67 DG |
2052 | ERR("Trying to check for data before version check"); |
2053 | ret = -1; | |
2054 | goto end_no_session; | |
2055 | } | |
2056 | ||
5312a3ed JG |
2057 | if (payload->size < sizeof(msg)) { |
2058 | ERR("Unexpected payload size in \"relay_begin_data_pending\": expected >= %zu bytes, got %zu bytes", | |
2059 | sizeof(msg), payload->size); | |
f7079f67 DG |
2060 | ret = -1; |
2061 | goto end_no_session; | |
2062 | } | |
5312a3ed JG |
2063 | memcpy(&msg, payload->data, sizeof(msg)); |
2064 | msg.session_id = be64toh(msg.session_id); | |
f7079f67 DG |
2065 | |
2066 | /* | |
7591bab1 MD |
2067 | * Iterate over all streams to set the begin data pending flag. |
2068 | * For now, the streams are indexed by stream handle so we have | |
2069 | * to iterate over all streams to find the one associated with | |
2070 | * the right session_id. | |
f7079f67 DG |
2071 | */ |
2072 | rcu_read_lock(); | |
d3e2ba59 | 2073 | cds_lfht_for_each_entry(relay_streams_ht->ht, &iter.iter, stream, |
2a174661 | 2074 | node.node) { |
7591bab1 MD |
2075 | if (!stream_get(stream)) { |
2076 | continue; | |
2077 | } | |
5312a3ed | 2078 | if (stream->trace->session->id == msg.session_id) { |
7591bab1 MD |
2079 | pthread_mutex_lock(&stream->lock); |
2080 | stream->data_pending_check_done = false; | |
2081 | pthread_mutex_unlock(&stream->lock); | |
f7079f67 DG |
2082 | DBG("Set begin data pending flag to stream %" PRIu64, |
2083 | stream->stream_handle); | |
2084 | } | |
7591bab1 | 2085 | stream_put(stream); |
f7079f67 DG |
2086 | } |
2087 | rcu_read_unlock(); | |
2088 | ||
53efb85a | 2089 | memset(&reply, 0, sizeof(reply)); |
f7079f67 DG |
2090 | /* All good, send back reply. */ |
2091 | reply.ret_code = htobe32(LTTNG_OK); | |
2092 | ||
5312a3ed JG |
2093 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
2094 | if (send_ret < (ssize_t) sizeof(reply)) { | |
2095 | ERR("Failed to send \"begin data pending\" command reply (ret = %zd)", | |
2096 | send_ret); | |
2097 | ret = -1; | |
2098 | } else { | |
2099 | ret = 0; | |
f7079f67 DG |
2100 | } |
2101 | ||
2102 | end_no_session: | |
2103 | return ret; | |
2104 | } | |
2105 | ||
2106 | /* | |
7591bab1 MD |
2107 | * End data pending command. This will check, for a given session id, if |
2108 | * each stream associated with it has its data_pending_check_done flag | |
2109 | * set. If not, this means that the client lost track of the stream but | |
2110 | * the data is still being streamed on our side. In this case, we inform | |
2111 | * the client that data is in flight. | |
f7079f67 DG |
2112 | * |
2113 | * Return to the client if there is data in flight or not with a ret_code. | |
2114 | */ | |
5312a3ed JG |
2115 | static int relay_end_data_pending(const struct lttcomm_relayd_hdr *recv_hdr, |
2116 | struct relay_connection *conn, | |
2117 | const struct lttng_buffer_view *payload) | |
f7079f67 DG |
2118 | { |
2119 | int ret; | |
5312a3ed | 2120 | ssize_t send_ret; |
f7079f67 DG |
2121 | struct lttng_ht_iter iter; |
2122 | struct lttcomm_relayd_end_data_pending msg; | |
2123 | struct lttcomm_relayd_generic_reply reply; | |
2124 | struct relay_stream *stream; | |
f7079f67 DG |
2125 | uint32_t is_data_inflight = 0; |
2126 | ||
f7079f67 DG |
2127 | DBG("End data pending command"); |
2128 | ||
5312a3ed | 2129 | if (!conn->session || !conn->version_check_done) { |
f7079f67 DG |
2130 | ERR("Trying to check for data before version check"); |
2131 | ret = -1; | |
2132 | goto end_no_session; | |
2133 | } | |
2134 | ||
5312a3ed JG |
2135 | if (payload->size < sizeof(msg)) { |
2136 | ERR("Unexpected payload size in \"relay_end_data_pending\": expected >= %zu bytes, got %zu bytes", | |
2137 | sizeof(msg), payload->size); | |
f7079f67 DG |
2138 | ret = -1; |
2139 | goto end_no_session; | |
2140 | } | |
5312a3ed JG |
2141 | memcpy(&msg, payload->data, sizeof(msg)); |
2142 | msg.session_id = be64toh(msg.session_id); | |
f7079f67 | 2143 | |
7591bab1 MD |
2144 | /* |
2145 | * Iterate over all streams to see if the begin data pending | |
2146 | * flag is set. | |
2147 | */ | |
f7079f67 | 2148 | rcu_read_lock(); |
d3e2ba59 | 2149 | cds_lfht_for_each_entry(relay_streams_ht->ht, &iter.iter, stream, |
2a174661 | 2150 | node.node) { |
7591bab1 MD |
2151 | if (!stream_get(stream)) { |
2152 | continue; | |
2153 | } | |
5312a3ed | 2154 | if (stream->trace->session->id != msg.session_id) { |
7591bab1 MD |
2155 | stream_put(stream); |
2156 | continue; | |
2157 | } | |
2158 | pthread_mutex_lock(&stream->lock); | |
2159 | if (!stream->data_pending_check_done) { | |
298a25ca JG |
2160 | uint64_t stream_seq; |
2161 | ||
2162 | if (session_streams_have_index(conn->session)) { | |
2163 | /* | |
2164 | * Ensure that both the index and stream data have been | |
2165 | * flushed up to the requested point. | |
2166 | */ | |
a8f9f353 | 2167 | stream_seq = min(stream->prev_data_seq, stream->prev_index_seq); |
298a25ca | 2168 | } else { |
a8f9f353 | 2169 | stream_seq = stream->prev_data_seq; |
298a25ca JG |
2170 | } |
2171 | if (!stream->closed || !(((int64_t) (stream_seq - stream->last_net_seq_num)) >= 0)) { | |
7591bab1 MD |
2172 | is_data_inflight = 1; |
2173 | DBG("Data is still in flight for stream %" PRIu64, | |
2174 | stream->stream_handle); | |
2175 | pthread_mutex_unlock(&stream->lock); | |
2176 | stream_put(stream); | |
2177 | break; | |
2178 | } | |
f7079f67 | 2179 | } |
7591bab1 MD |
2180 | pthread_mutex_unlock(&stream->lock); |
2181 | stream_put(stream); | |
f7079f67 DG |
2182 | } |
2183 | rcu_read_unlock(); | |
2184 | ||
53efb85a | 2185 | memset(&reply, 0, sizeof(reply)); |
f7079f67 DG |
2186 | /* All good, send back reply. */ |
2187 | reply.ret_code = htobe32(is_data_inflight); | |
2188 | ||
5312a3ed JG |
2189 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
2190 | if (send_ret < (ssize_t) sizeof(reply)) { | |
2191 | ERR("Failed to send \"end data pending\" command reply (ret = %zd)", | |
2192 | send_ret); | |
2193 | ret = -1; | |
2194 | } else { | |
2195 | ret = 0; | |
f7079f67 DG |
2196 | } |
2197 | ||
2198 | end_no_session: | |
2199 | return ret; | |
2200 | } | |
2201 | ||
1c20f0e2 JD |
2202 | /* |
2203 | * Receive an index for a specific stream. | |
2204 | * | |
2205 | * Return 0 on success else a negative value. | |
2206 | */ | |
5312a3ed JG |
2207 | static int relay_recv_index(const struct lttcomm_relayd_hdr *recv_hdr, |
2208 | struct relay_connection *conn, | |
2209 | const struct lttng_buffer_view *payload) | |
1c20f0e2 | 2210 | { |
5312a3ed JG |
2211 | int ret; |
2212 | ssize_t send_ret; | |
58eb9381 | 2213 | struct relay_session *session = conn->session; |
1c20f0e2 | 2214 | struct lttcomm_relayd_index index_info; |
1c20f0e2 JD |
2215 | struct lttcomm_relayd_generic_reply reply; |
2216 | struct relay_stream *stream; | |
f8f3885c | 2217 | size_t msg_len; |
1c20f0e2 | 2218 | |
58eb9381 | 2219 | assert(conn); |
1c20f0e2 JD |
2220 | |
2221 | DBG("Relay receiving index"); | |
2222 | ||
5312a3ed | 2223 | if (!session || !conn->version_check_done) { |
1c20f0e2 JD |
2224 | ERR("Trying to close a stream before version check"); |
2225 | ret = -1; | |
2226 | goto end_no_session; | |
2227 | } | |
2228 | ||
f8f3885c MD |
2229 | msg_len = lttcomm_relayd_index_len( |
2230 | lttng_to_index_major(conn->major, conn->minor), | |
2231 | lttng_to_index_minor(conn->major, conn->minor)); | |
5312a3ed JG |
2232 | if (payload->size < msg_len) { |
2233 | ERR("Unexpected payload size in \"relay_recv_index\": expected >= %zu bytes, got %zu bytes", | |
2234 | msg_len, payload->size); | |
1c20f0e2 JD |
2235 | ret = -1; |
2236 | goto end_no_session; | |
2237 | } | |
5312a3ed JG |
2238 | memcpy(&index_info, payload->data, msg_len); |
2239 | index_info.relay_stream_id = be64toh(index_info.relay_stream_id); | |
2240 | index_info.net_seq_num = be64toh(index_info.net_seq_num); | |
2241 | index_info.packet_size = be64toh(index_info.packet_size); | |
2242 | index_info.content_size = be64toh(index_info.content_size); | |
2243 | index_info.timestamp_begin = be64toh(index_info.timestamp_begin); | |
2244 | index_info.timestamp_end = be64toh(index_info.timestamp_end); | |
2245 | index_info.events_discarded = be64toh(index_info.events_discarded); | |
2246 | index_info.stream_id = be64toh(index_info.stream_id); | |
81df238b JR |
2247 | |
2248 | if (conn->minor >= 8) { | |
2249 | index_info.stream_instance_id = | |
2250 | be64toh(index_info.stream_instance_id); | |
2251 | index_info.packet_seq_num = be64toh(index_info.packet_seq_num); | |
0f83d1cc MD |
2252 | } else { |
2253 | index_info.stream_instance_id = -1ULL; | |
2254 | index_info.packet_seq_num = -1ULL; | |
81df238b | 2255 | } |
5312a3ed JG |
2256 | |
2257 | stream = stream_get_by_id(index_info.relay_stream_id); | |
1c20f0e2 | 2258 | if (!stream) { |
7591bab1 | 2259 | ERR("stream_get_by_id not found"); |
1c20f0e2 | 2260 | ret = -1; |
7591bab1 | 2261 | goto end; |
1c20f0e2 | 2262 | } |
d3e2ba59 | 2263 | |
c35f9726 JG |
2264 | pthread_mutex_lock(&stream->lock); |
2265 | ret = stream_add_index(stream, &index_info); | |
2266 | pthread_mutex_unlock(&stream->lock); | |
2267 | if (ret) { | |
7591bab1 MD |
2268 | goto end_stream_put; |
2269 | } | |
1c20f0e2 | 2270 | |
7591bab1 | 2271 | end_stream_put: |
7591bab1 | 2272 | stream_put(stream); |
7591bab1 | 2273 | end: |
53efb85a | 2274 | memset(&reply, 0, sizeof(reply)); |
1c20f0e2 JD |
2275 | if (ret < 0) { |
2276 | reply.ret_code = htobe32(LTTNG_ERR_UNK); | |
2277 | } else { | |
2278 | reply.ret_code = htobe32(LTTNG_OK); | |
2279 | } | |
58eb9381 | 2280 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
5312a3ed JG |
2281 | if (send_ret < (ssize_t) sizeof(reply)) { |
2282 | ERR("Failed to send \"recv index\" command reply (ret = %zd)", send_ret); | |
2283 | ret = -1; | |
1c20f0e2 JD |
2284 | } |
2285 | ||
2286 | end_no_session: | |
2287 | return ret; | |
2288 | } | |
2289 | ||
a4baae1b JD |
2290 | /* |
2291 | * Receive the streams_sent message. | |
2292 | * | |
2293 | * Return 0 on success else a negative value. | |
2294 | */ | |
5312a3ed JG |
2295 | static int relay_streams_sent(const struct lttcomm_relayd_hdr *recv_hdr, |
2296 | struct relay_connection *conn, | |
2297 | const struct lttng_buffer_view *payload) | |
a4baae1b | 2298 | { |
5312a3ed JG |
2299 | int ret; |
2300 | ssize_t send_ret; | |
a4baae1b JD |
2301 | struct lttcomm_relayd_generic_reply reply; |
2302 | ||
58eb9381 | 2303 | assert(conn); |
a4baae1b JD |
2304 | |
2305 | DBG("Relay receiving streams_sent"); | |
2306 | ||
5312a3ed | 2307 | if (!conn->session || !conn->version_check_done) { |
a4baae1b JD |
2308 | ERR("Trying to close a stream before version check"); |
2309 | ret = -1; | |
2310 | goto end_no_session; | |
2311 | } | |
2312 | ||
2313 | /* | |
7591bab1 MD |
2314 | * Publish every pending stream in the connection recv list which are |
2315 | * now ready to be used by the viewer. | |
4a9daf17 | 2316 | */ |
7591bab1 | 2317 | publish_connection_local_streams(conn); |
4a9daf17 | 2318 | |
53efb85a | 2319 | memset(&reply, 0, sizeof(reply)); |
a4baae1b | 2320 | reply.ret_code = htobe32(LTTNG_OK); |
58eb9381 | 2321 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
5312a3ed JG |
2322 | if (send_ret < (ssize_t) sizeof(reply)) { |
2323 | ERR("Failed to send \"streams sent\" command reply (ret = %zd)", | |
2324 | send_ret); | |
2325 | ret = -1; | |
a4baae1b JD |
2326 | } else { |
2327 | /* Success. */ | |
2328 | ret = 0; | |
2329 | } | |
2330 | ||
2331 | end_no_session: | |
2332 | return ret; | |
2333 | } | |
2334 | ||
d3ecc550 | 2335 | /* |
c35f9726 JG |
2336 | * relay_rotate_session_stream: rotate a stream to a new tracefile for the |
2337 | * session rotation feature (not the tracefile rotation feature). | |
d3ecc550 | 2338 | */ |
c35f9726 JG |
2339 | static int relay_rotate_session_streams( |
2340 | const struct lttcomm_relayd_hdr *recv_hdr, | |
5312a3ed JG |
2341 | struct relay_connection *conn, |
2342 | const struct lttng_buffer_view *payload) | |
d3ecc550 | 2343 | { |
30b9d5ab | 2344 | int ret = 0; |
c35f9726 | 2345 | uint32_t i; |
5312a3ed | 2346 | ssize_t send_ret; |
c35f9726 | 2347 | enum lttng_error_code reply_code = LTTNG_ERR_UNK; |
d3ecc550 | 2348 | struct relay_session *session = conn->session; |
c35f9726 JG |
2349 | struct lttcomm_relayd_rotate_streams rotate_streams; |
2350 | struct lttcomm_relayd_generic_reply reply = {}; | |
2351 | struct relay_stream *stream = NULL; | |
2352 | const size_t header_len = sizeof(struct lttcomm_relayd_rotate_streams); | |
2353 | struct lttng_trace_chunk *next_trace_chunk = NULL; | |
2354 | struct lttng_buffer_view stream_positions; | |
70626904 JG |
2355 | char chunk_id_buf[MAX_INT_DEC_LEN(uint64_t)]; |
2356 | const char *chunk_id_str = "none"; | |
d3ecc550 | 2357 | |
d3ecc550 JD |
2358 | if (!session || !conn->version_check_done) { |
2359 | ERR("Trying to rotate a stream before version check"); | |
2360 | ret = -1; | |
2361 | goto end_no_reply; | |
2362 | } | |
2363 | ||
2364 | if (session->major == 2 && session->minor < 11) { | |
2365 | ERR("Unsupported feature before 2.11"); | |
2366 | ret = -1; | |
2367 | goto end_no_reply; | |
2368 | } | |
2369 | ||
5312a3ed JG |
2370 | if (payload->size < header_len) { |
2371 | ERR("Unexpected payload size in \"relay_rotate_session_stream\": expected >= %zu bytes, got %zu bytes", | |
2372 | header_len, payload->size); | |
d3ecc550 JD |
2373 | ret = -1; |
2374 | goto end_no_reply; | |
2375 | } | |
2376 | ||
c35f9726 | 2377 | memcpy(&rotate_streams, payload->data, header_len); |
5312a3ed | 2378 | |
c35f9726 JG |
2379 | /* Convert header to host endianness. */ |
2380 | rotate_streams = (typeof(rotate_streams)) { | |
2381 | .stream_count = be32toh(rotate_streams.stream_count), | |
2382 | .new_chunk_id = (typeof(rotate_streams.new_chunk_id)) { | |
2383 | .is_set = !!rotate_streams.new_chunk_id.is_set, | |
2384 | .value = be64toh(rotate_streams.new_chunk_id.value), | |
2385 | } | |
2386 | }; | |
d3ecc550 | 2387 | |
c35f9726 JG |
2388 | if (rotate_streams.new_chunk_id.is_set) { |
2389 | /* | |
2390 | * Retrieve the trace chunk the stream must transition to. As | |
2391 | * per the protocol, this chunk should have been created | |
2392 | * before this command is received. | |
2393 | */ | |
2394 | next_trace_chunk = sessiond_trace_chunk_registry_get_chunk( | |
2395 | sessiond_trace_chunk_registry, | |
2396 | session->sessiond_uuid, session->id, | |
2397 | rotate_streams.new_chunk_id.value); | |
2398 | if (!next_trace_chunk) { | |
c70636a7 | 2399 | char uuid_str[LTTNG_UUID_STR_LEN]; |
c35f9726 JG |
2400 | |
2401 | lttng_uuid_to_str(session->sessiond_uuid, uuid_str); | |
2402 | ERR("Unknown next trace chunk in ROTATE_STREAMS command: sessiond_uuid = {%s}, session_id = %" PRIu64 | |
2403 | ", trace_chunk_id = %" PRIu64, | |
2404 | uuid_str, session->id, | |
2405 | rotate_streams.new_chunk_id.value); | |
2406 | reply_code = LTTNG_ERR_INVALID_PROTOCOL; | |
2407 | ret = -1; | |
2408 | goto end; | |
2409 | } | |
70626904 JG |
2410 | |
2411 | ret = snprintf(chunk_id_buf, sizeof(chunk_id_buf), "%" PRIu64, | |
2412 | rotate_streams.new_chunk_id.value); | |
2413 | if (ret < 0 || ret >= sizeof(chunk_id_buf)) { | |
2414 | chunk_id_str = "formatting error"; | |
2415 | } else { | |
2416 | chunk_id_str = chunk_id_buf; | |
2417 | } | |
d3ecc550 JD |
2418 | } |
2419 | ||
70626904 JG |
2420 | DBG("Rotate %" PRIu32 " streams of session \"%s\" to chunk \"%s\"", |
2421 | rotate_streams.stream_count, session->session_name, | |
2422 | chunk_id_str); | |
2423 | ||
c35f9726 JG |
2424 | stream_positions = lttng_buffer_view_from_view(payload, |
2425 | sizeof(rotate_streams), -1); | |
2426 | if (!stream_positions.data || | |
2427 | stream_positions.size < | |
2428 | (rotate_streams.stream_count * | |
2429 | sizeof(struct lttcomm_relayd_stream_rotation_position))) { | |
2430 | reply_code = LTTNG_ERR_INVALID_PROTOCOL; | |
d3ecc550 | 2431 | ret = -1; |
5312a3ed | 2432 | goto end; |
d3ecc550 JD |
2433 | } |
2434 | ||
c35f9726 JG |
2435 | for (i = 0; i < rotate_streams.stream_count; i++) { |
2436 | struct lttcomm_relayd_stream_rotation_position *position_comm = | |
2437 | &((typeof(position_comm)) stream_positions.data)[i]; | |
2438 | const struct lttcomm_relayd_stream_rotation_position pos = { | |
2439 | .stream_id = be64toh(position_comm->stream_id), | |
2440 | .rotate_at_seq_num = be64toh( | |
2441 | position_comm->rotate_at_seq_num), | |
2442 | }; | |
5312a3ed | 2443 | |
c35f9726 JG |
2444 | stream = stream_get_by_id(pos.stream_id); |
2445 | if (!stream) { | |
2446 | reply_code = LTTNG_ERR_INVALID; | |
2447 | ret = -1; | |
2448 | goto end; | |
c6db3843 JG |
2449 | } |
2450 | ||
c35f9726 JG |
2451 | pthread_mutex_lock(&stream->lock); |
2452 | ret = stream_set_pending_rotation(stream, next_trace_chunk, | |
2453 | pos.rotate_at_seq_num); | |
2454 | pthread_mutex_unlock(&stream->lock); | |
2455 | if (ret) { | |
2456 | reply_code = LTTNG_ERR_FILE_CREATION_ERROR; | |
2457 | goto end; | |
c6db3843 | 2458 | } |
c35f9726 JG |
2459 | |
2460 | stream_put(stream); | |
2461 | stream = NULL; | |
d3ecc550 JD |
2462 | } |
2463 | ||
c35f9726 | 2464 | reply_code = LTTNG_OK; |
eaeb64a9 | 2465 | ret = 0; |
d3ecc550 | 2466 | end: |
c35f9726 JG |
2467 | if (stream) { |
2468 | stream_put(stream); | |
d3ecc550 | 2469 | } |
c35f9726 JG |
2470 | |
2471 | reply.ret_code = htobe32((uint32_t) reply_code); | |
d3ecc550 JD |
2472 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, |
2473 | sizeof(struct lttcomm_relayd_generic_reply), 0); | |
5312a3ed JG |
2474 | if (send_ret < (ssize_t) sizeof(reply)) { |
2475 | ERR("Failed to send \"rotate session stream\" command reply (ret = %zd)", | |
2476 | send_ret); | |
2477 | ret = -1; | |
d3ecc550 | 2478 | } |
d3ecc550 | 2479 | end_no_reply: |
c35f9726 | 2480 | lttng_trace_chunk_put(next_trace_chunk); |
d3ecc550 JD |
2481 | return ret; |
2482 | } | |
2483 | ||
e5add6d0 | 2484 | |
e5add6d0 JG |
2485 | |
2486 | /* | |
2487 | * relay_create_trace_chunk: create a new trace chunk | |
2488 | */ | |
2489 | static int relay_create_trace_chunk(const struct lttcomm_relayd_hdr *recv_hdr, | |
2490 | struct relay_connection *conn, | |
2491 | const struct lttng_buffer_view *payload) | |
2492 | { | |
2493 | int ret = 0; | |
2494 | ssize_t send_ret; | |
2495 | struct relay_session *session = conn->session; | |
2496 | struct lttcomm_relayd_create_trace_chunk *msg; | |
2497 | struct lttcomm_relayd_generic_reply reply = {}; | |
2498 | struct lttng_buffer_view header_view; | |
2499 | struct lttng_buffer_view chunk_name_view; | |
2500 | struct lttng_trace_chunk *chunk = NULL, *published_chunk = NULL; | |
2501 | enum lttng_error_code reply_code = LTTNG_OK; | |
2502 | enum lttng_trace_chunk_status chunk_status; | |
cbf53d23 | 2503 | struct lttng_directory_handle *session_output = NULL; |
a7ceb342 | 2504 | const char *new_path; |
e5add6d0 JG |
2505 | |
2506 | if (!session || !conn->version_check_done) { | |
2507 | ERR("Trying to create a trace chunk before version check"); | |
2508 | ret = -1; | |
2509 | goto end_no_reply; | |
2510 | } | |
2511 | ||
2512 | if (session->major == 2 && session->minor < 11) { | |
2513 | ERR("Chunk creation command is unsupported before 2.11"); | |
2514 | ret = -1; | |
2515 | goto end_no_reply; | |
2516 | } | |
2517 | ||
2518 | header_view = lttng_buffer_view_from_view(payload, 0, sizeof(*msg)); | |
2519 | if (!header_view.data) { | |
2520 | ERR("Failed to receive payload of chunk creation command"); | |
2521 | ret = -1; | |
2522 | goto end_no_reply; | |
2523 | } | |
2524 | ||
2525 | /* Convert to host endianness. */ | |
2526 | msg = (typeof(msg)) header_view.data; | |
2527 | msg->chunk_id = be64toh(msg->chunk_id); | |
2528 | msg->creation_timestamp = be64toh(msg->creation_timestamp); | |
2529 | msg->override_name_length = be32toh(msg->override_name_length); | |
2530 | ||
a7ceb342 MD |
2531 | if (session->current_trace_chunk && |
2532 | !lttng_trace_chunk_get_name_overridden(session->current_trace_chunk)) { | |
2533 | chunk_status = lttng_trace_chunk_rename_path(session->current_trace_chunk, | |
2534 | DEFAULT_CHUNK_TMP_OLD_DIRECTORY); | |
2535 | if (chunk_status != LTTNG_TRACE_CHUNK_STATUS_OK) { | |
2536 | ERR("Failed to rename old chunk"); | |
2537 | ret = -1; | |
2538 | reply_code = LTTNG_ERR_UNK; | |
2539 | goto end; | |
2540 | } | |
2541 | } | |
2542 | session->ongoing_rotation = true; | |
2543 | if (!session->current_trace_chunk) { | |
2544 | if (!session->has_rotated) { | |
2545 | new_path = ""; | |
2546 | } else { | |
2547 | new_path = NULL; | |
2548 | } | |
2549 | } else { | |
2550 | new_path = DEFAULT_CHUNK_TMP_NEW_DIRECTORY; | |
2551 | } | |
e5add6d0 | 2552 | chunk = lttng_trace_chunk_create( |
a7ceb342 | 2553 | msg->chunk_id, msg->creation_timestamp, new_path); |
e5add6d0 JG |
2554 | if (!chunk) { |
2555 | ERR("Failed to create trace chunk in trace chunk creation command"); | |
2556 | ret = -1; | |
2557 | reply_code = LTTNG_ERR_NOMEM; | |
2558 | goto end; | |
2559 | } | |
2560 | ||
2561 | if (msg->override_name_length) { | |
2562 | const char *name; | |
2563 | ||
2564 | chunk_name_view = lttng_buffer_view_from_view(payload, | |
2565 | sizeof(*msg), | |
2566 | msg->override_name_length); | |
2567 | name = chunk_name_view.data; | |
2568 | if (!name || name[msg->override_name_length - 1]) { | |
2569 | ERR("Failed to receive payload of chunk creation command"); | |
2570 | ret = -1; | |
2571 | reply_code = LTTNG_ERR_INVALID; | |
2572 | goto end; | |
2573 | } | |
2574 | ||
2575 | chunk_status = lttng_trace_chunk_override_name( | |
2576 | chunk, chunk_name_view.data); | |
2577 | switch (chunk_status) { | |
2578 | case LTTNG_TRACE_CHUNK_STATUS_OK: | |
2579 | break; | |
2580 | case LTTNG_TRACE_CHUNK_STATUS_INVALID_ARGUMENT: | |
2581 | ERR("Failed to set the name of new trace chunk in trace chunk creation command (invalid name)"); | |
2582 | reply_code = LTTNG_ERR_INVALID; | |
2583 | ret = -1; | |
2584 | goto end; | |
2585 | default: | |
2586 | ERR("Failed to set the name of new trace chunk in trace chunk creation command (unknown error)"); | |
2587 | reply_code = LTTNG_ERR_UNK; | |
2588 | ret = -1; | |
2589 | goto end; | |
2590 | } | |
2591 | } | |
2592 | ||
e5add6d0 JG |
2593 | chunk_status = lttng_trace_chunk_set_credentials_current_user(chunk); |
2594 | if (chunk_status != LTTNG_TRACE_CHUNK_STATUS_OK) { | |
2595 | reply_code = LTTNG_ERR_UNK; | |
2596 | ret = -1; | |
2597 | goto end; | |
2598 | } | |
2599 | ||
cbf53d23 JG |
2600 | session_output = session_create_output_directory_handle( |
2601 | conn->session); | |
2602 | if (!session_output) { | |
0ccc0411 JG |
2603 | reply_code = LTTNG_ERR_CREATE_DIR_FAIL; |
2604 | goto end; | |
2605 | } | |
cbf53d23 JG |
2606 | chunk_status = lttng_trace_chunk_set_as_owner(chunk, session_output); |
2607 | lttng_directory_handle_put(session_output); | |
2608 | session_output = NULL; | |
e5add6d0 JG |
2609 | if (chunk_status != LTTNG_TRACE_CHUNK_STATUS_OK) { |
2610 | reply_code = LTTNG_ERR_UNK; | |
2611 | ret = -1; | |
2612 | goto end; | |
2613 | } | |
2614 | ||
2615 | published_chunk = sessiond_trace_chunk_registry_publish_chunk( | |
2616 | sessiond_trace_chunk_registry, | |
2617 | conn->session->sessiond_uuid, | |
2618 | conn->session->id, | |
2619 | chunk); | |
2620 | if (!published_chunk) { | |
c70636a7 | 2621 | char uuid_str[LTTNG_UUID_STR_LEN]; |
e5add6d0 JG |
2622 | |
2623 | lttng_uuid_to_str(conn->session->sessiond_uuid, uuid_str); | |
2624 | ERR("Failed to publish chunk: sessiond_uuid = %s, session_id = %" PRIu64 ", chunk_id = %" PRIu64, | |
2625 | uuid_str, | |
2626 | conn->session->id, | |
2627 | msg->chunk_id); | |
2628 | ret = -1; | |
2629 | reply_code = LTTNG_ERR_NOMEM; | |
2630 | goto end; | |
2631 | } | |
2632 | ||
2633 | pthread_mutex_lock(&conn->session->lock); | |
62bad3bf JG |
2634 | if (conn->session->pending_closure_trace_chunk) { |
2635 | /* | |
2636 | * Invalid; this means a second create_trace_chunk command was | |
2637 | * received before a close_trace_chunk. | |
2638 | */ | |
2639 | ERR("Invalid trace chunk close command received; a trace chunk is already waiting for a trace chunk close command"); | |
2640 | reply_code = LTTNG_ERR_INVALID_PROTOCOL; | |
2641 | ret = -1; | |
2642 | goto end_unlock_session; | |
2643 | } | |
2644 | conn->session->pending_closure_trace_chunk = | |
2645 | conn->session->current_trace_chunk; | |
e5add6d0 | 2646 | conn->session->current_trace_chunk = published_chunk; |
e5add6d0 | 2647 | published_chunk = NULL; |
a7ceb342 MD |
2648 | if (!conn->session->pending_closure_trace_chunk) { |
2649 | session->ongoing_rotation = false; | |
2650 | } | |
62bad3bf | 2651 | end_unlock_session: |
c35f9726 | 2652 | pthread_mutex_unlock(&conn->session->lock); |
e5add6d0 JG |
2653 | end: |
2654 | reply.ret_code = htobe32((uint32_t) reply_code); | |
2655 | send_ret = conn->sock->ops->sendmsg(conn->sock, | |
2656 | &reply, | |
2657 | sizeof(struct lttcomm_relayd_generic_reply), | |
2658 | 0); | |
2659 | if (send_ret < (ssize_t) sizeof(reply)) { | |
2660 | ERR("Failed to send \"create trace chunk\" command reply (ret = %zd)", | |
2661 | send_ret); | |
2662 | ret = -1; | |
2663 | } | |
2664 | end_no_reply: | |
2665 | lttng_trace_chunk_put(chunk); | |
2666 | lttng_trace_chunk_put(published_chunk); | |
cbf53d23 | 2667 | lttng_directory_handle_put(session_output); |
e5add6d0 JG |
2668 | return ret; |
2669 | } | |
2670 | ||
bbc4768c JG |
2671 | /* |
2672 | * relay_close_trace_chunk: close a trace chunk | |
2673 | */ | |
2674 | static int relay_close_trace_chunk(const struct lttcomm_relayd_hdr *recv_hdr, | |
2675 | struct relay_connection *conn, | |
2676 | const struct lttng_buffer_view *payload) | |
2677 | { | |
9898f786 | 2678 | int ret = 0, buf_ret; |
bbc4768c JG |
2679 | ssize_t send_ret; |
2680 | struct relay_session *session = conn->session; | |
2681 | struct lttcomm_relayd_close_trace_chunk *msg; | |
ecd1a12f | 2682 | struct lttcomm_relayd_close_trace_chunk_reply reply = {}; |
bbc4768c JG |
2683 | struct lttng_buffer_view header_view; |
2684 | struct lttng_trace_chunk *chunk = NULL; | |
2685 | enum lttng_error_code reply_code = LTTNG_OK; | |
2686 | enum lttng_trace_chunk_status chunk_status; | |
2687 | uint64_t chunk_id; | |
c35f9726 | 2688 | LTTNG_OPTIONAL(enum lttng_trace_chunk_command_type) close_command = {}; |
bbc4768c | 2689 | time_t close_timestamp; |
ecd1a12f MD |
2690 | char closed_trace_chunk_path[LTTNG_PATH_MAX]; |
2691 | size_t path_length = 0; | |
2692 | const char *chunk_name = NULL; | |
2693 | struct lttng_dynamic_buffer reply_payload; | |
a7ceb342 | 2694 | const char *new_path; |
ecd1a12f MD |
2695 | |
2696 | lttng_dynamic_buffer_init(&reply_payload); | |
bbc4768c JG |
2697 | |
2698 | if (!session || !conn->version_check_done) { | |
2699 | ERR("Trying to close a trace chunk before version check"); | |
2700 | ret = -1; | |
2701 | goto end_no_reply; | |
2702 | } | |
2703 | ||
2704 | if (session->major == 2 && session->minor < 11) { | |
2705 | ERR("Chunk close command is unsupported before 2.11"); | |
2706 | ret = -1; | |
2707 | goto end_no_reply; | |
2708 | } | |
2709 | ||
2710 | header_view = lttng_buffer_view_from_view(payload, 0, sizeof(*msg)); | |
2711 | if (!header_view.data) { | |
2712 | ERR("Failed to receive payload of chunk close command"); | |
2713 | ret = -1; | |
2714 | goto end_no_reply; | |
2715 | } | |
2716 | ||
2717 | /* Convert to host endianness. */ | |
2718 | msg = (typeof(msg)) header_view.data; | |
2719 | chunk_id = be64toh(msg->chunk_id); | |
2720 | close_timestamp = (time_t) be64toh(msg->close_timestamp); | |
2721 | close_command = (typeof(close_command)){ | |
2722 | .value = be32toh(msg->close_command.value), | |
2723 | .is_set = msg->close_command.is_set, | |
2724 | }; | |
2725 | ||
2726 | chunk = sessiond_trace_chunk_registry_get_chunk( | |
2727 | sessiond_trace_chunk_registry, | |
2728 | conn->session->sessiond_uuid, | |
2729 | conn->session->id, | |
2730 | chunk_id); | |
2731 | if (!chunk) { | |
c70636a7 | 2732 | char uuid_str[LTTNG_UUID_STR_LEN]; |
bbc4768c JG |
2733 | |
2734 | lttng_uuid_to_str(conn->session->sessiond_uuid, uuid_str); | |
2735 | ERR("Failed to find chunk to close: sessiond_uuid = %s, session_id = %" PRIu64 ", chunk_id = %" PRIu64, | |
2736 | uuid_str, | |
2737 | conn->session->id, | |
2738 | msg->chunk_id); | |
2739 | ret = -1; | |
2740 | reply_code = LTTNG_ERR_NOMEM; | |
2741 | goto end; | |
2742 | } | |
2743 | ||
62bad3bf | 2744 | pthread_mutex_lock(&session->lock); |
f77a6ec2 MD |
2745 | if (close_command.is_set && |
2746 | close_command.value == LTTNG_TRACE_CHUNK_COMMAND_TYPE_DELETE) { | |
2747 | /* | |
2748 | * Clear command. It is a protocol error to ask for a | |
2749 | * clear on a relay which does not allow it. Querying | |
2750 | * the configuration allows figuring out whether | |
2751 | * clearing is allowed before doing the clear. | |
2752 | */ | |
2753 | if (!opt_allow_clear) { | |
2754 | ret = -1; | |
2755 | reply_code = LTTNG_ERR_INVALID_PROTOCOL; | |
2756 | goto end_unlock_session; | |
2757 | } | |
2758 | } | |
62bad3bf JG |
2759 | if (session->pending_closure_trace_chunk && |
2760 | session->pending_closure_trace_chunk != chunk) { | |
2761 | ERR("Trace chunk close command for session \"%s\" does not target the trace chunk pending closure", | |
2762 | session->session_name); | |
2763 | reply_code = LTTNG_ERR_INVALID_PROTOCOL; | |
2764 | ret = -1; | |
2765 | goto end_unlock_session; | |
2766 | } | |
2767 | ||
a7ceb342 MD |
2768 | if (session->current_trace_chunk && session->current_trace_chunk != chunk && |
2769 | !lttng_trace_chunk_get_name_overridden(session->current_trace_chunk)) { | |
2770 | if (close_command.is_set && | |
2771 | close_command.value == LTTNG_TRACE_CHUNK_COMMAND_TYPE_DELETE && | |
2772 | !session->has_rotated) { | |
2773 | /* New chunk stays in session output directory. */ | |
2774 | new_path = ""; | |
2775 | } else { | |
2776 | /* Use chunk name for new chunk. */ | |
2777 | new_path = NULL; | |
2778 | } | |
2779 | /* Rename new chunk path. */ | |
2780 | chunk_status = lttng_trace_chunk_rename_path(session->current_trace_chunk, | |
2781 | new_path); | |
2782 | if (chunk_status != LTTNG_TRACE_CHUNK_STATUS_OK) { | |
2783 | ret = -1; | |
2784 | goto end; | |
2785 | } | |
2786 | session->ongoing_rotation = false; | |
2787 | } | |
2788 | if ((!close_command.is_set || | |
2789 | close_command.value == LTTNG_TRACE_CHUNK_COMMAND_TYPE_NO_OPERATION) && | |
2790 | !lttng_trace_chunk_get_name_overridden(chunk)) { | |
2791 | const char *old_path; | |
2792 | ||
2793 | if (!session->has_rotated) { | |
2794 | old_path = ""; | |
2795 | } else { | |
2796 | old_path = NULL; | |
2797 | } | |
2798 | /* We need to move back the .tmp_old_chunk to its rightful place. */ | |
2799 | chunk_status = lttng_trace_chunk_rename_path(chunk, old_path); | |
2800 | if (chunk_status != LTTNG_TRACE_CHUNK_STATUS_OK) { | |
2801 | ret = -1; | |
2802 | goto end; | |
2803 | } | |
2804 | } | |
bbc4768c JG |
2805 | chunk_status = lttng_trace_chunk_set_close_timestamp( |
2806 | chunk, close_timestamp); | |
2807 | if (chunk_status != LTTNG_TRACE_CHUNK_STATUS_OK) { | |
2808 | ERR("Failed to set trace chunk close timestamp"); | |
2809 | ret = -1; | |
2810 | reply_code = LTTNG_ERR_UNK; | |
62bad3bf | 2811 | goto end_unlock_session; |
bbc4768c JG |
2812 | } |
2813 | ||
2814 | if (close_command.is_set) { | |
2815 | chunk_status = lttng_trace_chunk_set_close_command( | |
2816 | chunk, close_command.value); | |
2817 | if (chunk_status != LTTNG_TRACE_CHUNK_STATUS_OK) { | |
2818 | ret = -1; | |
2819 | reply_code = LTTNG_ERR_INVALID; | |
62bad3bf | 2820 | goto end_unlock_session; |
bbc4768c JG |
2821 | } |
2822 | } | |
ecd1a12f MD |
2823 | chunk_status = lttng_trace_chunk_get_name(chunk, &chunk_name, NULL); |
2824 | if (chunk_status != LTTNG_TRACE_CHUNK_STATUS_OK) { | |
2825 | ERR("Failed to get chunk name"); | |
2826 | ret = -1; | |
2827 | reply_code = LTTNG_ERR_UNK; | |
2828 | goto end_unlock_session; | |
2829 | } | |
2830 | if (!session->has_rotated && !session->snapshot) { | |
2831 | ret = lttng_strncpy(closed_trace_chunk_path, | |
2832 | session->output_path, | |
2833 | sizeof(closed_trace_chunk_path)); | |
2834 | if (ret) { | |
2835 | ERR("Failed to send trace chunk path: path length of %zu bytes exceeds the maximal allowed length of %zu bytes", | |
2836 | strlen(session->output_path), | |
2837 | sizeof(closed_trace_chunk_path)); | |
2838 | reply_code = LTTNG_ERR_NOMEM; | |
2839 | ret = -1; | |
2840 | goto end_unlock_session; | |
2841 | } | |
2842 | } else { | |
2843 | if (session->snapshot) { | |
2844 | ret = snprintf(closed_trace_chunk_path, | |
2845 | sizeof(closed_trace_chunk_path), | |
2846 | "%s/%s", session->output_path, | |
2847 | chunk_name); | |
2848 | } else { | |
2849 | ret = snprintf(closed_trace_chunk_path, | |
2850 | sizeof(closed_trace_chunk_path), | |
2851 | "%s/" DEFAULT_ARCHIVED_TRACE_CHUNKS_DIRECTORY | |
2852 | "/%s", | |
2853 | session->output_path, chunk_name); | |
2854 | } | |
2855 | if (ret < 0 || ret == sizeof(closed_trace_chunk_path)) { | |
2856 | ERR("Failed to format closed trace chunk resulting path"); | |
2857 | reply_code = ret < 0 ? LTTNG_ERR_UNK : LTTNG_ERR_NOMEM; | |
2858 | ret = -1; | |
2859 | goto end_unlock_session; | |
2860 | } | |
2861 | } | |
489ea154 MD |
2862 | if (close_command.is_set && |
2863 | close_command.value == LTTNG_TRACE_CHUNK_COMMAND_TYPE_MOVE_TO_COMPLETED) { | |
2864 | session->has_rotated = true; | |
2865 | } | |
ecd1a12f MD |
2866 | DBG("Reply chunk path on close: %s", closed_trace_chunk_path); |
2867 | path_length = strlen(closed_trace_chunk_path) + 1; | |
2868 | if (path_length > UINT32_MAX) { | |
2869 | ERR("Closed trace chunk path exceeds the maximal length allowed by the protocol"); | |
2870 | ret = -1; | |
2871 | reply_code = LTTNG_ERR_INVALID_PROTOCOL; | |
2872 | goto end_unlock_session; | |
2873 | } | |
bbc4768c | 2874 | |
c35f9726 JG |
2875 | if (session->current_trace_chunk == chunk) { |
2876 | /* | |
2877 | * After a trace chunk close command, no new streams | |
2878 | * referencing the chunk may be created. Hence, on the | |
2879 | * event that no new trace chunk have been created for | |
2880 | * the session, the reference to the current trace chunk | |
2881 | * is released in order to allow it to be reclaimed when | |
2882 | * the last stream releases its reference to it. | |
2883 | */ | |
2884 | lttng_trace_chunk_put(session->current_trace_chunk); | |
2885 | session->current_trace_chunk = NULL; | |
2886 | } | |
62bad3bf JG |
2887 | lttng_trace_chunk_put(session->pending_closure_trace_chunk); |
2888 | session->pending_closure_trace_chunk = NULL; | |
2889 | end_unlock_session: | |
c35f9726 JG |
2890 | pthread_mutex_unlock(&session->lock); |
2891 | ||
bbc4768c | 2892 | end: |
ecd1a12f MD |
2893 | reply.generic.ret_code = htobe32((uint32_t) reply_code); |
2894 | reply.path_length = htobe32((uint32_t) path_length); | |
9898f786 | 2895 | buf_ret = lttng_dynamic_buffer_append( |
ecd1a12f | 2896 | &reply_payload, &reply, sizeof(reply)); |
9898f786 | 2897 | if (buf_ret) { |
ecd1a12f MD |
2898 | ERR("Failed to append \"close trace chunk\" command reply header to payload buffer"); |
2899 | goto end_no_reply; | |
2900 | } | |
2901 | ||
2902 | if (reply_code == LTTNG_OK) { | |
9898f786 | 2903 | buf_ret = lttng_dynamic_buffer_append(&reply_payload, |
ecd1a12f | 2904 | closed_trace_chunk_path, path_length); |
9898f786 | 2905 | if (buf_ret) { |
ecd1a12f MD |
2906 | ERR("Failed to append \"close trace chunk\" command reply path to payload buffer"); |
2907 | goto end_no_reply; | |
2908 | } | |
2909 | } | |
2910 | ||
bbc4768c | 2911 | send_ret = conn->sock->ops->sendmsg(conn->sock, |
ecd1a12f MD |
2912 | reply_payload.data, |
2913 | reply_payload.size, | |
bbc4768c | 2914 | 0); |
ecd1a12f MD |
2915 | if (send_ret < reply_payload.size) { |
2916 | ERR("Failed to send \"close trace chunk\" command reply of %zu bytes (ret = %zd)", | |
2917 | reply_payload.size, send_ret); | |
bbc4768c | 2918 | ret = -1; |
ecd1a12f | 2919 | goto end_no_reply; |
bbc4768c JG |
2920 | } |
2921 | end_no_reply: | |
2922 | lttng_trace_chunk_put(chunk); | |
ecd1a12f | 2923 | lttng_dynamic_buffer_reset(&reply_payload); |
bbc4768c JG |
2924 | return ret; |
2925 | } | |
2926 | ||
c35f9726 JG |
2927 | /* |
2928 | * relay_trace_chunk_exists: check if a trace chunk exists | |
2929 | */ | |
2930 | static int relay_trace_chunk_exists(const struct lttcomm_relayd_hdr *recv_hdr, | |
2931 | struct relay_connection *conn, | |
2932 | const struct lttng_buffer_view *payload) | |
2933 | { | |
2934 | int ret = 0; | |
2935 | ssize_t send_ret; | |
2936 | struct relay_session *session = conn->session; | |
2937 | struct lttcomm_relayd_trace_chunk_exists *msg; | |
2938 | struct lttcomm_relayd_trace_chunk_exists_reply reply = {}; | |
2939 | struct lttng_buffer_view header_view; | |
c35f9726 | 2940 | uint64_t chunk_id; |
6b584c2e | 2941 | bool chunk_exists; |
c35f9726 JG |
2942 | |
2943 | if (!session || !conn->version_check_done) { | |
2944 | ERR("Trying to close a trace chunk before version check"); | |
2945 | ret = -1; | |
2946 | goto end_no_reply; | |
2947 | } | |
2948 | ||
2949 | if (session->major == 2 && session->minor < 11) { | |
2950 | ERR("Chunk close command is unsupported before 2.11"); | |
2951 | ret = -1; | |
2952 | goto end_no_reply; | |
2953 | } | |
2954 | ||
2955 | header_view = lttng_buffer_view_from_view(payload, 0, sizeof(*msg)); | |
2956 | if (!header_view.data) { | |
2957 | ERR("Failed to receive payload of chunk close command"); | |
2958 | ret = -1; | |
2959 | goto end_no_reply; | |
2960 | } | |
2961 | ||
2962 | /* Convert to host endianness. */ | |
2963 | msg = (typeof(msg)) header_view.data; | |
2964 | chunk_id = be64toh(msg->chunk_id); | |
2965 | ||
6b584c2e | 2966 | ret = sessiond_trace_chunk_registry_chunk_exists( |
c35f9726 JG |
2967 | sessiond_trace_chunk_registry, |
2968 | conn->session->sessiond_uuid, | |
2969 | conn->session->id, | |
6b584c2e JG |
2970 | chunk_id, &chunk_exists); |
2971 | /* | |
2972 | * If ret is not 0, send the reply and report the error to the caller. | |
2973 | * It is a protocol (or internal) error and the session/connection | |
2974 | * should be torn down. | |
2975 | */ | |
2976 | reply = (typeof(reply)){ | |
2977 | .generic.ret_code = htobe32((uint32_t) | |
2978 | (ret == 0 ? LTTNG_OK : LTTNG_ERR_INVALID_PROTOCOL)), | |
2979 | .trace_chunk_exists = ret == 0 ? chunk_exists : 0, | |
c35f9726 | 2980 | }; |
6b584c2e JG |
2981 | send_ret = conn->sock->ops->sendmsg( |
2982 | conn->sock, &reply, sizeof(reply), 0); | |
c35f9726 JG |
2983 | if (send_ret < (ssize_t) sizeof(reply)) { |
2984 | ERR("Failed to send \"create trace chunk\" command reply (ret = %zd)", | |
2985 | send_ret); | |
2986 | ret = -1; | |
2987 | } | |
2988 | end_no_reply: | |
c35f9726 JG |
2989 | return ret; |
2990 | } | |
2991 | ||
8614e600 MD |
2992 | /* |
2993 | * relay_get_configuration: query whether feature is available | |
2994 | */ | |
2995 | static int relay_get_configuration(const struct lttcomm_relayd_hdr *recv_hdr, | |
2996 | struct relay_connection *conn, | |
2997 | const struct lttng_buffer_view *payload) | |
2998 | { | |
2999 | int ret = 0; | |
3000 | ssize_t send_ret; | |
3001 | struct lttcomm_relayd_get_configuration *msg; | |
3002 | struct lttcomm_relayd_get_configuration_reply reply = {}; | |
3003 | struct lttng_buffer_view header_view; | |
3004 | uint64_t query_flags = 0; | |
3005 | uint64_t result_flags = 0; | |
3006 | ||
3007 | header_view = lttng_buffer_view_from_view(payload, 0, sizeof(*msg)); | |
3008 | if (!header_view.data) { | |
3009 | ERR("Failed to receive payload of chunk close command"); | |
3010 | ret = -1; | |
3011 | goto end_no_reply; | |
3012 | } | |
3013 | ||
3014 | /* Convert to host endianness. */ | |
3015 | msg = (typeof(msg)) header_view.data; | |
3016 | query_flags = be64toh(msg->query_flags); | |
3017 | ||
3018 | if (query_flags) { | |
3019 | ret = LTTNG_ERR_INVALID_PROTOCOL; | |
3020 | goto reply; | |
3021 | } | |
3022 | if (opt_allow_clear) { | |
3023 | result_flags |= LTTCOMM_RELAYD_CONFIGURATION_FLAG_CLEAR_ALLOWED; | |
3024 | } | |
3025 | ret = 0; | |
3026 | reply: | |
3027 | reply = (typeof(reply)){ | |
3028 | .generic.ret_code = htobe32((uint32_t) | |
3029 | (ret == 0 ? LTTNG_OK : LTTNG_ERR_INVALID_PROTOCOL)), | |
3030 | .relayd_configuration_flags = htobe64(result_flags), | |
3031 | }; | |
3032 | send_ret = conn->sock->ops->sendmsg( | |
3033 | conn->sock, &reply, sizeof(reply), 0); | |
3034 | if (send_ret < (ssize_t) sizeof(reply)) { | |
3035 | ERR("Failed to send \"get configuration\" command reply (ret = %zd)", | |
3036 | send_ret); | |
3037 | ret = -1; | |
3038 | } | |
3039 | end_no_reply: | |
3040 | return ret; | |
3041 | } | |
3042 | ||
5312a3ed JG |
3043 | #define DBG_CMD(cmd_name, conn) \ |
3044 | DBG3("Processing \"%s\" command for socket %i", cmd_name, conn->sock->fd); | |
3045 | ||
3046 | static int relay_process_control_command(struct relay_connection *conn, | |
3047 | const struct lttcomm_relayd_hdr *header, | |
3048 | const struct lttng_buffer_view *payload) | |
b8aa1682 JD |
3049 | { |
3050 | int ret = 0; | |
3051 | ||
5312a3ed | 3052 | switch (header->cmd) { |
b8aa1682 | 3053 | case RELAYD_CREATE_SESSION: |
5312a3ed JG |
3054 | DBG_CMD("RELAYD_CREATE_SESSION", conn); |
3055 | ret = relay_create_session(header, conn, payload); | |
b8aa1682 | 3056 | break; |
b8aa1682 | 3057 | case RELAYD_ADD_STREAM: |
5312a3ed JG |
3058 | DBG_CMD("RELAYD_ADD_STREAM", conn); |
3059 | ret = relay_add_stream(header, conn, payload); | |
b8aa1682 JD |
3060 | break; |
3061 | case RELAYD_START_DATA: | |
5312a3ed JG |
3062 | DBG_CMD("RELAYD_START_DATA", conn); |
3063 | ret = relay_start(header, conn, payload); | |
b8aa1682 JD |
3064 | break; |
3065 | case RELAYD_SEND_METADATA: | |
5312a3ed JG |
3066 | DBG_CMD("RELAYD_SEND_METADATA", conn); |
3067 | ret = relay_recv_metadata(header, conn, payload); | |
b8aa1682 JD |
3068 | break; |
3069 | case RELAYD_VERSION: | |
5312a3ed JG |
3070 | DBG_CMD("RELAYD_VERSION", conn); |
3071 | ret = relay_send_version(header, conn, payload); | |
b8aa1682 | 3072 | break; |
173af62f | 3073 | case RELAYD_CLOSE_STREAM: |
5312a3ed JG |
3074 | DBG_CMD("RELAYD_CLOSE_STREAM", conn); |
3075 | ret = relay_close_stream(header, conn, payload); | |
173af62f | 3076 | break; |
6d805429 | 3077 | case RELAYD_DATA_PENDING: |
5312a3ed JG |
3078 | DBG_CMD("RELAYD_DATA_PENDING", conn); |
3079 | ret = relay_data_pending(header, conn, payload); | |
c8f59ee5 DG |
3080 | break; |
3081 | case RELAYD_QUIESCENT_CONTROL: | |
5312a3ed JG |
3082 | DBG_CMD("RELAYD_QUIESCENT_CONTROL", conn); |
3083 | ret = relay_quiescent_control(header, conn, payload); | |
c8f59ee5 | 3084 | break; |
f7079f67 | 3085 | case RELAYD_BEGIN_DATA_PENDING: |
5312a3ed JG |
3086 | DBG_CMD("RELAYD_BEGIN_DATA_PENDING", conn); |
3087 | ret = relay_begin_data_pending(header, conn, payload); | |
f7079f67 DG |
3088 | break; |
3089 | case RELAYD_END_DATA_PENDING: | |
5312a3ed JG |
3090 | DBG_CMD("RELAYD_END_DATA_PENDING", conn); |
3091 | ret = relay_end_data_pending(header, conn, payload); | |
f7079f67 | 3092 | break; |
1c20f0e2 | 3093 | case RELAYD_SEND_INDEX: |
5312a3ed JG |
3094 | DBG_CMD("RELAYD_SEND_INDEX", conn); |
3095 | ret = relay_recv_index(header, conn, payload); | |
1c20f0e2 | 3096 | break; |
a4baae1b | 3097 | case RELAYD_STREAMS_SENT: |
5312a3ed JG |
3098 | DBG_CMD("RELAYD_STREAMS_SENT", conn); |
3099 | ret = relay_streams_sent(header, conn, payload); | |
a4baae1b | 3100 | break; |
93ec662e | 3101 | case RELAYD_RESET_METADATA: |
5312a3ed JG |
3102 | DBG_CMD("RELAYD_RESET_METADATA", conn); |
3103 | ret = relay_reset_metadata(header, conn, payload); | |
93ec662e | 3104 | break; |
c35f9726 JG |
3105 | case RELAYD_ROTATE_STREAMS: |
3106 | DBG_CMD("RELAYD_ROTATE_STREAMS", conn); | |
3107 | ret = relay_rotate_session_streams(header, conn, payload); | |
d3ecc550 | 3108 | break; |
e5add6d0 JG |
3109 | case RELAYD_CREATE_TRACE_CHUNK: |
3110 | DBG_CMD("RELAYD_CREATE_TRACE_CHUNK", conn); | |
3111 | ret = relay_create_trace_chunk(header, conn, payload); | |
3112 | break; | |
bbc4768c JG |
3113 | case RELAYD_CLOSE_TRACE_CHUNK: |
3114 | DBG_CMD("RELAYD_CLOSE_TRACE_CHUNK", conn); | |
3115 | ret = relay_close_trace_chunk(header, conn, payload); | |
3116 | break; | |
c35f9726 JG |
3117 | case RELAYD_TRACE_CHUNK_EXISTS: |
3118 | DBG_CMD("RELAYD_TRACE_CHUNK_EXISTS", conn); | |
3119 | ret = relay_trace_chunk_exists(header, conn, payload); | |
3120 | break; | |
8614e600 MD |
3121 | case RELAYD_GET_CONFIGURATION: |
3122 | DBG_CMD("RELAYD_GET_CONFIGURATION", conn); | |
3123 | ret = relay_get_configuration(header, conn, payload); | |
3124 | break; | |
b8aa1682 JD |
3125 | case RELAYD_UPDATE_SYNC_INFO: |
3126 | default: | |
5312a3ed | 3127 | ERR("Received unknown command (%u)", header->cmd); |
58eb9381 | 3128 | relay_unknown_command(conn); |
b8aa1682 JD |
3129 | ret = -1; |
3130 | goto end; | |
3131 | } | |
3132 | ||
3133 | end: | |
3134 | return ret; | |
3135 | } | |
3136 | ||
5569b118 JG |
3137 | static enum relay_connection_status relay_process_control_receive_payload( |
3138 | struct relay_connection *conn) | |
5312a3ed JG |
3139 | { |
3140 | int ret = 0; | |
5569b118 | 3141 | enum relay_connection_status status = RELAY_CONNECTION_STATUS_OK; |
5312a3ed JG |
3142 | struct lttng_dynamic_buffer *reception_buffer = |
3143 | &conn->protocol.ctrl.reception_buffer; | |
3144 | struct ctrl_connection_state_receive_payload *state = | |
3145 | &conn->protocol.ctrl.state.receive_payload; | |
3146 | struct lttng_buffer_view payload_view; | |
3147 | ||
3148 | if (state->left_to_receive == 0) { | |
3149 | /* Short-circuit for payload-less commands. */ | |
3150 | goto reception_complete; | |
3151 | } | |
3152 | ||
3153 | ret = conn->sock->ops->recvmsg(conn->sock, | |
3154 | reception_buffer->data + state->received, | |
3155 | state->left_to_receive, MSG_DONTWAIT); | |
3156 | if (ret < 0) { | |
5569b118 JG |
3157 | if (errno != EAGAIN && errno != EWOULDBLOCK) { |
3158 | PERROR("Unable to receive command payload on sock %d", | |
3159 | conn->sock->fd); | |
3160 | status = RELAY_CONNECTION_STATUS_ERROR; | |
3161 | } | |
5312a3ed JG |
3162 | goto end; |
3163 | } else if (ret == 0) { | |
3164 | DBG("Socket %d performed an orderly shutdown (received EOF)", conn->sock->fd); | |
5569b118 | 3165 | status = RELAY_CONNECTION_STATUS_CLOSED; |
5312a3ed JG |
3166 | goto end; |
3167 | } | |
3168 | ||
3169 | assert(ret > 0); | |
3170 | assert(ret <= state->left_to_receive); | |
3171 | ||
3172 | state->left_to_receive -= ret; | |
3173 | state->received += ret; | |
3174 | ||
3175 | if (state->left_to_receive > 0) { | |
3176 | /* | |
3177 | * Can't transition to the protocol's next state, wait to | |
3178 | * receive the rest of the header. | |
3179 | */ | |
3180 | DBG3("Partial reception of control connection protocol payload (received %" PRIu64 " bytes, %" PRIu64 " bytes left to receive, fd = %i)", | |
3181 | state->received, state->left_to_receive, | |
3182 | conn->sock->fd); | |
5312a3ed JG |
3183 | goto end; |
3184 | } | |
3185 | ||
3186 | reception_complete: | |
3187 | DBG("Done receiving control command payload: fd = %i, payload size = %" PRIu64 " bytes", | |
3188 | conn->sock->fd, state->received); | |
3189 | /* | |
3190 | * The payload required to process the command has been received. | |
3191 | * A view to the reception buffer is forwarded to the various | |
3192 | * commands and the state of the control is reset on success. | |
3193 | * | |
3194 | * Commands are responsible for sending their reply to the peer. | |
3195 | */ | |
3196 | payload_view = lttng_buffer_view_from_dynamic_buffer(reception_buffer, | |
3197 | 0, -1); | |
3198 | ret = relay_process_control_command(conn, | |
3199 | &state->header, &payload_view); | |
3200 | if (ret < 0) { | |
5569b118 | 3201 | status = RELAY_CONNECTION_STATUS_ERROR; |
5312a3ed JG |
3202 | goto end; |
3203 | } | |
3204 | ||
3205 | ret = connection_reset_protocol_state(conn); | |
5569b118 JG |
3206 | if (ret) { |
3207 | status = RELAY_CONNECTION_STATUS_ERROR; | |
3208 | } | |
5312a3ed | 3209 | end: |
5569b118 | 3210 | return status; |
5312a3ed JG |
3211 | } |
3212 | ||
5569b118 JG |
3213 | static enum relay_connection_status relay_process_control_receive_header( |
3214 | struct relay_connection *conn) | |
5312a3ed JG |
3215 | { |
3216 | int ret = 0; | |
5569b118 | 3217 | enum relay_connection_status status = RELAY_CONNECTION_STATUS_OK; |
5312a3ed JG |
3218 | struct lttcomm_relayd_hdr header; |
3219 | struct lttng_dynamic_buffer *reception_buffer = | |
3220 | &conn->protocol.ctrl.reception_buffer; | |
3221 | struct ctrl_connection_state_receive_header *state = | |
3222 | &conn->protocol.ctrl.state.receive_header; | |
3223 | ||
3224 | assert(state->left_to_receive != 0); | |
3225 | ||
3226 | ret = conn->sock->ops->recvmsg(conn->sock, | |
3227 | reception_buffer->data + state->received, | |
3228 | state->left_to_receive, MSG_DONTWAIT); | |
3229 | if (ret < 0) { | |
5569b118 JG |
3230 | if (errno != EAGAIN && errno != EWOULDBLOCK) { |
3231 | PERROR("Unable to receive control command header on sock %d", | |
3232 | conn->sock->fd); | |
3233 | status = RELAY_CONNECTION_STATUS_ERROR; | |
3234 | } | |
5312a3ed JG |
3235 | goto end; |
3236 | } else if (ret == 0) { | |
3237 | DBG("Socket %d performed an orderly shutdown (received EOF)", conn->sock->fd); | |
5569b118 | 3238 | status = RELAY_CONNECTION_STATUS_CLOSED; |
5312a3ed JG |
3239 | goto end; |
3240 | } | |
3241 | ||
3242 | assert(ret > 0); | |
3243 | assert(ret <= state->left_to_receive); | |
3244 | ||
3245 | state->left_to_receive -= ret; | |
3246 | state->received += ret; | |
3247 | ||
3248 | if (state->left_to_receive > 0) { | |
3249 | /* | |
3250 | * Can't transition to the protocol's next state, wait to | |
3251 | * receive the rest of the header. | |
3252 | */ | |
3253 | DBG3("Partial reception of control connection protocol header (received %" PRIu64 " bytes, %" PRIu64 " bytes left to receive, fd = %i)", | |
3254 | state->received, state->left_to_receive, | |
3255 | conn->sock->fd); | |
5312a3ed JG |
3256 | goto end; |
3257 | } | |
3258 | ||
3259 | /* Transition to next state: receiving the command's payload. */ | |
3260 | conn->protocol.ctrl.state_id = | |
3261 | CTRL_CONNECTION_STATE_RECEIVE_PAYLOAD; | |
3262 | memcpy(&header, reception_buffer->data, sizeof(header)); | |
3263 | header.circuit_id = be64toh(header.circuit_id); | |
3264 | header.data_size = be64toh(header.data_size); | |
3265 | header.cmd = be32toh(header.cmd); | |
3266 | header.cmd_version = be32toh(header.cmd_version); | |
3267 | memcpy(&conn->protocol.ctrl.state.receive_payload.header, | |
3268 | &header, sizeof(header)); | |
3269 | ||
3270 | DBG("Done receiving control command header: fd = %i, cmd = %" PRIu32 ", cmd_version = %" PRIu32 ", payload size = %" PRIu64 " bytes", | |
3271 | conn->sock->fd, header.cmd, header.cmd_version, | |
3272 | header.data_size); | |
3273 | ||
715e6fb1 | 3274 | if (header.data_size > DEFAULT_NETWORK_RELAYD_CTRL_MAX_PAYLOAD_SIZE) { |
5312a3ed JG |
3275 | ERR("Command header indicates a payload (%" PRIu64 " bytes) that exceeds the maximal payload size allowed on a control connection.", |
3276 | header.data_size); | |
5569b118 | 3277 | status = RELAY_CONNECTION_STATUS_ERROR; |
5312a3ed JG |
3278 | goto end; |
3279 | } | |
3280 | ||
3281 | conn->protocol.ctrl.state.receive_payload.left_to_receive = | |
3282 | header.data_size; | |
3283 | conn->protocol.ctrl.state.receive_payload.received = 0; | |
3284 | ret = lttng_dynamic_buffer_set_size(reception_buffer, | |
3285 | header.data_size); | |
3286 | if (ret) { | |
5569b118 | 3287 | status = RELAY_CONNECTION_STATUS_ERROR; |
5312a3ed JG |
3288 | goto end; |
3289 | } | |
3290 | ||
3291 | if (header.data_size == 0) { | |
3292 | /* | |
3293 | * Manually invoke the next state as the poll loop | |
3294 | * will not wake-up to allow us to proceed further. | |
3295 | */ | |
5569b118 | 3296 | status = relay_process_control_receive_payload(conn); |
5312a3ed JG |
3297 | } |
3298 | end: | |
5569b118 | 3299 | return status; |
5312a3ed JG |
3300 | } |
3301 | ||
3302 | /* | |
3303 | * Process the commands received on the control socket | |
3304 | */ | |
5569b118 JG |
3305 | static enum relay_connection_status relay_process_control( |
3306 | struct relay_connection *conn) | |
5312a3ed | 3307 | { |
5569b118 | 3308 | enum relay_connection_status status; |
5312a3ed JG |
3309 | |
3310 | switch (conn->protocol.ctrl.state_id) { | |
3311 | case CTRL_CONNECTION_STATE_RECEIVE_HEADER: | |
5569b118 | 3312 | status = relay_process_control_receive_header(conn); |
5312a3ed JG |
3313 | break; |
3314 | case CTRL_CONNECTION_STATE_RECEIVE_PAYLOAD: | |
5569b118 | 3315 | status = relay_process_control_receive_payload(conn); |
5312a3ed JG |
3316 | break; |
3317 | default: | |
3318 | ERR("Unknown control connection protocol state encountered."); | |
3319 | abort(); | |
3320 | } | |
3321 | ||
5569b118 | 3322 | return status; |
5312a3ed JG |
3323 | } |
3324 | ||
5569b118 JG |
3325 | static enum relay_connection_status relay_process_data_receive_header( |
3326 | struct relay_connection *conn) | |
b8aa1682 | 3327 | { |
5312a3ed | 3328 | int ret; |
5569b118 | 3329 | enum relay_connection_status status = RELAY_CONNECTION_STATUS_OK; |
5312a3ed JG |
3330 | struct data_connection_state_receive_header *state = |
3331 | &conn->protocol.data.state.receive_header; | |
3332 | struct lttcomm_relayd_data_hdr header; | |
b8aa1682 | 3333 | struct relay_stream *stream; |
5312a3ed JG |
3334 | |
3335 | assert(state->left_to_receive != 0); | |
3336 | ||
3337 | ret = conn->sock->ops->recvmsg(conn->sock, | |
3338 | state->header_reception_buffer + state->received, | |
3339 | state->left_to_receive, MSG_DONTWAIT); | |
3340 | if (ret < 0) { | |
5569b118 JG |
3341 | if (errno != EAGAIN && errno != EWOULDBLOCK) { |
3342 | PERROR("Unable to receive data header on sock %d", conn->sock->fd); | |
3343 | status = RELAY_CONNECTION_STATUS_ERROR; | |
3344 | } | |
5312a3ed JG |
3345 | goto end; |
3346 | } else if (ret == 0) { | |
3347 | /* Orderly shutdown. Not necessary to print an error. */ | |
3348 | DBG("Socket %d performed an orderly shutdown (received EOF)", conn->sock->fd); | |
5569b118 | 3349 | status = RELAY_CONNECTION_STATUS_CLOSED; |
b8aa1682 JD |
3350 | goto end; |
3351 | } | |
3352 | ||
5312a3ed JG |
3353 | assert(ret > 0); |
3354 | assert(ret <= state->left_to_receive); | |
3355 | ||
3356 | state->left_to_receive -= ret; | |
3357 | state->received += ret; | |
3358 | ||
3359 | if (state->left_to_receive > 0) { | |
3360 | /* | |
3361 | * Can't transition to the protocol's next state, wait to | |
3362 | * receive the rest of the header. | |
3363 | */ | |
3364 | DBG3("Partial reception of data connection header (received %" PRIu64 " bytes, %" PRIu64 " bytes left to receive, fd = %i)", | |
3365 | state->received, state->left_to_receive, | |
3366 | conn->sock->fd); | |
7591bab1 | 3367 | goto end; |
b8aa1682 | 3368 | } |
b8aa1682 | 3369 | |
5312a3ed JG |
3370 | /* Transition to next state: receiving the payload. */ |
3371 | conn->protocol.data.state_id = DATA_CONNECTION_STATE_RECEIVE_PAYLOAD; | |
173af62f | 3372 | |
5312a3ed JG |
3373 | memcpy(&header, state->header_reception_buffer, sizeof(header)); |
3374 | header.circuit_id = be64toh(header.circuit_id); | |
3375 | header.stream_id = be64toh(header.stream_id); | |
3376 | header.data_size = be32toh(header.data_size); | |
3377 | header.net_seq_num = be64toh(header.net_seq_num); | |
3378 | header.padding_size = be32toh(header.padding_size); | |
3379 | memcpy(&conn->protocol.data.state.receive_payload.header, &header, sizeof(header)); | |
3380 | ||
3381 | conn->protocol.data.state.receive_payload.left_to_receive = | |
3382 | header.data_size; | |
3383 | conn->protocol.data.state.receive_payload.received = 0; | |
3384 | conn->protocol.data.state.receive_payload.rotate_index = false; | |
3385 | ||
3386 | DBG("Received data connection header on fd %i: circuit_id = %" PRIu64 ", stream_id = %" PRIu64 ", data_size = %" PRIu32 ", net_seq_num = %" PRIu64 ", padding_size = %" PRIu32, | |
3387 | conn->sock->fd, header.circuit_id, | |
3388 | header.stream_id, header.data_size, | |
3389 | header.net_seq_num, header.padding_size); | |
3390 | ||
3391 | stream = stream_get_by_id(header.stream_id); | |
3392 | if (!stream) { | |
3393 | DBG("relay_process_data_receive_payload: Cannot find stream %" PRIu64, | |
3394 | header.stream_id); | |
5569b118 JG |
3395 | /* Protocol error. */ |
3396 | status = RELAY_CONNECTION_STATUS_ERROR; | |
5312a3ed JG |
3397 | goto end; |
3398 | } | |
b8aa1682 | 3399 | |
7591bab1 | 3400 | pthread_mutex_lock(&stream->lock); |
c35f9726 JG |
3401 | /* Prepare stream for the reception of a new packet. */ |
3402 | ret = stream_init_packet(stream, header.data_size, | |
3403 | &conn->protocol.data.state.receive_payload.rotate_index); | |
3404 | pthread_mutex_unlock(&stream->lock); | |
3405 | if (ret) { | |
3406 | ERR("Failed to rotate stream output file"); | |
3407 | status = RELAY_CONNECTION_STATUS_ERROR; | |
3408 | goto end_stream_unlock; | |
1c20f0e2 JD |
3409 | } |
3410 | ||
5312a3ed | 3411 | end_stream_unlock: |
5312a3ed JG |
3412 | stream_put(stream); |
3413 | end: | |
5569b118 | 3414 | return status; |
5312a3ed JG |
3415 | } |
3416 | ||
5569b118 JG |
3417 | static enum relay_connection_status relay_process_data_receive_payload( |
3418 | struct relay_connection *conn) | |
5312a3ed JG |
3419 | { |
3420 | int ret; | |
5569b118 | 3421 | enum relay_connection_status status = RELAY_CONNECTION_STATUS_OK; |
5312a3ed JG |
3422 | struct relay_stream *stream; |
3423 | struct data_connection_state_receive_payload *state = | |
3424 | &conn->protocol.data.state.receive_payload; | |
3425 | const size_t chunk_size = RECV_DATA_BUFFER_SIZE; | |
3426 | char data_buffer[chunk_size]; | |
3427 | bool partial_recv = false; | |
3428 | bool new_stream = false, close_requested = false, index_flushed = false; | |
3429 | uint64_t left_to_receive = state->left_to_receive; | |
3430 | struct relay_session *session; | |
3431 | ||
fd0f1e3e JR |
3432 | DBG3("Receiving data for stream id %" PRIu64 " seqnum %" PRIu64 ", %" PRIu64" bytes received, %" PRIu64 " bytes left to receive", |
3433 | state->header.stream_id, state->header.net_seq_num, | |
3434 | state->received, left_to_receive); | |
3435 | ||
5312a3ed JG |
3436 | stream = stream_get_by_id(state->header.stream_id); |
3437 | if (!stream) { | |
5569b118 | 3438 | /* Protocol error. */ |
fd0f1e3e | 3439 | ERR("relay_process_data_receive_payload: cannot find stream %" PRIu64, |
5312a3ed | 3440 | state->header.stream_id); |
5569b118 | 3441 | status = RELAY_CONNECTION_STATUS_ERROR; |
5312a3ed | 3442 | goto end; |
1c20f0e2 JD |
3443 | } |
3444 | ||
5312a3ed JG |
3445 | pthread_mutex_lock(&stream->lock); |
3446 | session = stream->trace->session; | |
fd0f1e3e JR |
3447 | if (!conn->session) { |
3448 | ret = connection_set_session(conn, session); | |
3449 | if (ret) { | |
3450 | status = RELAY_CONNECTION_STATUS_ERROR; | |
3451 | goto end_stream_unlock; | |
3452 | } | |
3453 | } | |
5312a3ed JG |
3454 | |
3455 | /* | |
3456 | * The size of the "chunk" received on any iteration is bounded by: | |
3457 | * - the data left to receive, | |
3458 | * - the data immediately available on the socket, | |
3459 | * - the on-stack data buffer | |
3460 | */ | |
3461 | while (left_to_receive > 0 && !partial_recv) { | |
5312a3ed | 3462 | size_t recv_size = min(left_to_receive, chunk_size); |
c35f9726 | 3463 | struct lttng_buffer_view packet_chunk; |
5312a3ed JG |
3464 | |
3465 | ret = conn->sock->ops->recvmsg(conn->sock, data_buffer, | |
3466 | recv_size, MSG_DONTWAIT); | |
3467 | if (ret < 0) { | |
5569b118 JG |
3468 | if (errno != EAGAIN && errno != EWOULDBLOCK) { |
3469 | PERROR("Socket %d error", conn->sock->fd); | |
3470 | status = RELAY_CONNECTION_STATUS_ERROR; | |
3471 | } | |
0848dba7 | 3472 | goto end_stream_unlock; |
5312a3ed JG |
3473 | } else if (ret == 0) { |
3474 | /* No more data ready to be consumed on socket. */ | |
3475 | DBG3("No more data ready for consumption on data socket of stream id %" PRIu64, | |
3476 | state->header.stream_id); | |
5569b118 | 3477 | status = RELAY_CONNECTION_STATUS_CLOSED; |
5312a3ed JG |
3478 | break; |
3479 | } else if (ret < (int) recv_size) { | |
3480 | /* | |
3481 | * All the data available on the socket has been | |
3482 | * consumed. | |
3483 | */ | |
3484 | partial_recv = true; | |
c35f9726 | 3485 | recv_size = ret; |
0848dba7 MD |
3486 | } |
3487 | ||
c35f9726 JG |
3488 | packet_chunk = lttng_buffer_view_init(data_buffer, |
3489 | 0, recv_size); | |
3490 | assert(packet_chunk.data); | |
5312a3ed | 3491 | |
c35f9726 JG |
3492 | ret = stream_write(stream, &packet_chunk, 0); |
3493 | if (ret) { | |
0848dba7 | 3494 | ERR("Relay error writing data to file"); |
5569b118 | 3495 | status = RELAY_CONNECTION_STATUS_ERROR; |
0848dba7 MD |
3496 | goto end_stream_unlock; |
3497 | } | |
3498 | ||
5312a3ed JG |
3499 | left_to_receive -= recv_size; |
3500 | state->received += recv_size; | |
3501 | state->left_to_receive = left_to_receive; | |
5312a3ed JG |
3502 | } |
3503 | ||
3504 | if (state->left_to_receive > 0) { | |
3505 | /* | |
3506 | * Did not receive all the data expected, wait for more data to | |
3507 | * become available on the socket. | |
3508 | */ | |
3509 | DBG3("Partial receive on data connection of stream id %" PRIu64 ", %" PRIu64 " bytes received, %" PRIu64 " bytes left to receive", | |
3510 | state->header.stream_id, state->received, | |
3511 | state->left_to_receive); | |
5312a3ed | 3512 | goto end_stream_unlock; |
0848dba7 | 3513 | } |
5ab7344e | 3514 | |
c35f9726 JG |
3515 | ret = stream_write(stream, NULL, state->header.padding_size); |
3516 | if (ret) { | |
5569b118 | 3517 | status = RELAY_CONNECTION_STATUS_ERROR; |
7591bab1 | 3518 | goto end_stream_unlock; |
1d4dfdef | 3519 | } |
5312a3ed | 3520 | |
298a25ca | 3521 | if (session_streams_have_index(session)) { |
c35f9726 JG |
3522 | ret = stream_update_index(stream, state->header.net_seq_num, |
3523 | state->rotate_index, &index_flushed, | |
3524 | state->header.data_size + state->header.padding_size); | |
5312a3ed | 3525 | if (ret < 0) { |
c35f9726 | 3526 | ERR("Failed to update index: stream %" PRIu64 " net_seq_num %" PRIu64 " ret %d", |
5312a3ed JG |
3527 | stream->stream_handle, |
3528 | state->header.net_seq_num, ret); | |
5569b118 | 3529 | status = RELAY_CONNECTION_STATUS_ERROR; |
5312a3ed JG |
3530 | goto end_stream_unlock; |
3531 | } | |
3532 | } | |
3533 | ||
a8f9f353 | 3534 | if (stream->prev_data_seq == -1ULL) { |
c0bae11d MD |
3535 | new_stream = true; |
3536 | } | |
3537 | ||
c35f9726 JG |
3538 | ret = stream_complete_packet(stream, state->header.data_size + |
3539 | state->header.padding_size, state->header.net_seq_num, | |
3540 | index_flushed); | |
3541 | if (ret) { | |
3542 | status = RELAY_CONNECTION_STATUS_ERROR; | |
3543 | goto end_stream_unlock; | |
3544 | } | |
5312a3ed JG |
3545 | |
3546 | /* | |
3547 | * Resetting the protocol state (to RECEIVE_HEADER) will trash the | |
3548 | * contents of *state which are aliased (union) to the same location as | |
3549 | * the new state. Don't use it beyond this point. | |
3550 | */ | |
3551 | connection_reset_protocol_state(conn); | |
3552 | state = NULL; | |
173af62f | 3553 | |
7591bab1 | 3554 | end_stream_unlock: |
bda7c7b9 | 3555 | close_requested = stream->close_requested; |
7591bab1 | 3556 | pthread_mutex_unlock(&stream->lock); |
5312a3ed | 3557 | if (close_requested && left_to_receive == 0) { |
bda7c7b9 JG |
3558 | try_stream_close(stream); |
3559 | } | |
3560 | ||
c0bae11d MD |
3561 | if (new_stream) { |
3562 | pthread_mutex_lock(&session->lock); | |
3563 | uatomic_set(&session->new_streams, 1); | |
3564 | pthread_mutex_unlock(&session->lock); | |
3565 | } | |
5312a3ed | 3566 | |
7591bab1 | 3567 | stream_put(stream); |
b8aa1682 | 3568 | end: |
5569b118 | 3569 | return status; |
b8aa1682 JD |
3570 | } |
3571 | ||
5312a3ed JG |
3572 | /* |
3573 | * relay_process_data: Process the data received on the data socket | |
3574 | */ | |
5569b118 JG |
3575 | static enum relay_connection_status relay_process_data( |
3576 | struct relay_connection *conn) | |
5312a3ed | 3577 | { |
5569b118 | 3578 | enum relay_connection_status status; |
5312a3ed JG |
3579 | |
3580 | switch (conn->protocol.data.state_id) { | |
3581 | case DATA_CONNECTION_STATE_RECEIVE_HEADER: | |
5569b118 | 3582 | status = relay_process_data_receive_header(conn); |
5312a3ed JG |
3583 | break; |
3584 | case DATA_CONNECTION_STATE_RECEIVE_PAYLOAD: | |
5569b118 | 3585 | status = relay_process_data_receive_payload(conn); |
5312a3ed JG |
3586 | break; |
3587 | default: | |
3588 | ERR("Unexpected data connection communication state."); | |
3589 | abort(); | |
3590 | } | |
3591 | ||
5569b118 | 3592 | return status; |
5312a3ed JG |
3593 | } |
3594 | ||
7591bab1 | 3595 | static void cleanup_connection_pollfd(struct lttng_poll_event *events, int pollfd) |
b8aa1682 JD |
3596 | { |
3597 | int ret; | |
3598 | ||
58eb9381 | 3599 | (void) lttng_poll_del(events, pollfd); |
b8aa1682 JD |
3600 | |
3601 | ret = close(pollfd); | |
3602 | if (ret < 0) { | |
3603 | ERR("Closing pollfd %d", pollfd); | |
3604 | } | |
3605 | } | |
3606 | ||
7591bab1 MD |
3607 | static void relay_thread_close_connection(struct lttng_poll_event *events, |
3608 | int pollfd, struct relay_connection *conn) | |
9d1bbf21 | 3609 | { |
7591bab1 | 3610 | const char *type_str; |
2a174661 | 3611 | |
7591bab1 MD |
3612 | switch (conn->type) { |
3613 | case RELAY_DATA: | |
3614 | type_str = "Data"; | |
3615 | break; | |
3616 | case RELAY_CONTROL: | |
3617 | type_str = "Control"; | |
3618 | break; | |
3619 | case RELAY_VIEWER_COMMAND: | |
3620 | type_str = "Viewer Command"; | |
3621 | break; | |
3622 | case RELAY_VIEWER_NOTIFICATION: | |
3623 | type_str = "Viewer Notification"; | |
3624 | break; | |
3625 | default: | |
3626 | type_str = "Unknown"; | |
9d1bbf21 | 3627 | } |
7591bab1 MD |
3628 | cleanup_connection_pollfd(events, pollfd); |
3629 | connection_put(conn); | |
3630 | DBG("%s connection closed with %d", type_str, pollfd); | |
b8aa1682 JD |
3631 | } |
3632 | ||
3633 | /* | |
3634 | * This thread does the actual work | |
3635 | */ | |
7591bab1 | 3636 | static void *relay_thread_worker(void *data) |
b8aa1682 | 3637 | { |
beaad64c DG |
3638 | int ret, err = -1, last_seen_data_fd = -1; |
3639 | uint32_t nb_fd; | |
b8aa1682 JD |
3640 | struct lttng_poll_event events; |
3641 | struct lttng_ht *relay_connections_ht; | |
b8aa1682 | 3642 | struct lttng_ht_iter iter; |
90e7d72f | 3643 | struct relay_connection *destroy_conn = NULL; |
b8aa1682 JD |
3644 | |
3645 | DBG("[thread] Relay worker started"); | |
3646 | ||
9d1bbf21 MD |
3647 | rcu_register_thread(); |
3648 | ||
55706a7d MD |
3649 | health_register(health_relayd, HEALTH_RELAYD_TYPE_WORKER); |
3650 | ||
9b5e0863 MD |
3651 | if (testpoint(relayd_thread_worker)) { |
3652 | goto error_testpoint; | |
3653 | } | |
3654 | ||
f385ae0a MD |
3655 | health_code_update(); |
3656 | ||
b8aa1682 JD |
3657 | /* table of connections indexed on socket */ |
3658 | relay_connections_ht = lttng_ht_new(0, LTTNG_HT_TYPE_ULONG); | |
095a4ae5 MD |
3659 | if (!relay_connections_ht) { |
3660 | goto relay_connections_ht_error; | |
3661 | } | |
b8aa1682 | 3662 | |
b8aa1682 JD |
3663 | ret = create_thread_poll_set(&events, 2); |
3664 | if (ret < 0) { | |
3665 | goto error_poll_create; | |
3666 | } | |
3667 | ||
58eb9381 | 3668 | ret = lttng_poll_add(&events, relay_conn_pipe[0], LPOLLIN | LPOLLRDHUP); |
b8aa1682 JD |
3669 | if (ret < 0) { |
3670 | goto error; | |
3671 | } | |
3672 | ||
beaad64c | 3673 | restart: |
b8aa1682 | 3674 | while (1) { |
beaad64c DG |
3675 | int idx = -1, i, seen_control = 0, last_notdel_data_fd = -1; |
3676 | ||
f385ae0a MD |
3677 | health_code_update(); |
3678 | ||
b8aa1682 | 3679 | /* Infinite blocking call, waiting for transmission */ |
87c1611d | 3680 | DBG3("Relayd worker thread polling..."); |
f385ae0a | 3681 | health_poll_entry(); |
b8aa1682 | 3682 | ret = lttng_poll_wait(&events, -1); |
f385ae0a | 3683 | health_poll_exit(); |
b8aa1682 JD |
3684 | if (ret < 0) { |
3685 | /* | |
3686 | * Restart interrupted system call. | |
3687 | */ | |
3688 | if (errno == EINTR) { | |
3689 | goto restart; | |
3690 | } | |
3691 | goto error; | |
3692 | } | |
3693 | ||
0d9c5d77 DG |
3694 | nb_fd = ret; |
3695 | ||
beaad64c | 3696 | /* |
7591bab1 MD |
3697 | * Process control. The control connection is |
3698 | * prioritized so we don't starve it with high | |
3699 | * throughput tracing data on the data connection. | |
beaad64c | 3700 | */ |
b8aa1682 JD |
3701 | for (i = 0; i < nb_fd; i++) { |
3702 | /* Fetch once the poll data */ | |
beaad64c DG |
3703 | uint32_t revents = LTTNG_POLL_GETEV(&events, i); |
3704 | int pollfd = LTTNG_POLL_GETFD(&events, i); | |
b8aa1682 | 3705 | |
f385ae0a MD |
3706 | health_code_update(); |
3707 | ||
b8aa1682 JD |
3708 | /* Thread quit pipe has been closed. Killing thread. */ |
3709 | ret = check_thread_quit_pipe(pollfd, revents); | |
3710 | if (ret) { | |
095a4ae5 MD |
3711 | err = 0; |
3712 | goto exit; | |
b8aa1682 JD |
3713 | } |
3714 | ||
58eb9381 DG |
3715 | /* Inspect the relay conn pipe for new connection */ |
3716 | if (pollfd == relay_conn_pipe[0]) { | |
03e43155 | 3717 | if (revents & LPOLLIN) { |
90e7d72f JG |
3718 | struct relay_connection *conn; |
3719 | ||
58eb9381 | 3720 | ret = lttng_read(relay_conn_pipe[0], &conn, sizeof(conn)); |
b8aa1682 JD |
3721 | if (ret < 0) { |
3722 | goto error; | |
3723 | } | |
73039936 FD |
3724 | ret = lttng_poll_add(&events, |
3725 | conn->sock->fd, | |
58eb9381 | 3726 | LPOLLIN | LPOLLRDHUP); |
73039936 FD |
3727 | if (ret) { |
3728 | ERR("Failed to add new connection file descriptor to poll set"); | |
3729 | goto error; | |
3730 | } | |
7591bab1 | 3731 | connection_ht_add(relay_connections_ht, conn); |
58eb9381 | 3732 | DBG("Connection socket %d added", conn->sock->fd); |
03e43155 MD |
3733 | } else if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) { |
3734 | ERR("Relay connection pipe error"); | |
3735 | goto error; | |
3736 | } else { | |
3737 | ERR("Unexpected poll events %u for sock %d", revents, pollfd); | |
3738 | goto error; | |
b8aa1682 | 3739 | } |
58eb9381 | 3740 | } else { |
90e7d72f JG |
3741 | struct relay_connection *ctrl_conn; |
3742 | ||
7591bab1 | 3743 | ctrl_conn = connection_get_by_sock(relay_connections_ht, pollfd); |
58eb9381 | 3744 | /* If not found, there is a synchronization issue. */ |
90e7d72f | 3745 | assert(ctrl_conn); |
58eb9381 | 3746 | |
03e43155 MD |
3747 | if (ctrl_conn->type == RELAY_DATA) { |
3748 | if (revents & LPOLLIN) { | |
beaad64c DG |
3749 | /* |
3750 | * Flag the last seen data fd not deleted. It will be | |
3751 | * used as the last seen fd if any fd gets deleted in | |
3752 | * this first loop. | |
3753 | */ | |
3754 | last_notdel_data_fd = pollfd; | |
3755 | } | |
03e43155 MD |
3756 | goto put_ctrl_connection; |
3757 | } | |
3758 | assert(ctrl_conn->type == RELAY_CONTROL); | |
3759 | ||
3760 | if (revents & LPOLLIN) { | |
5569b118 JG |
3761 | enum relay_connection_status status; |
3762 | ||
3763 | status = relay_process_control(ctrl_conn); | |
3764 | if (status != RELAY_CONNECTION_STATUS_OK) { | |
fd0f1e3e JR |
3765 | /* |
3766 | * On socket error flag the session as aborted to force | |
3767 | * the cleanup of its stream otherwise it can leak | |
3768 | * during the lifetime of the relayd. | |
3769 | * | |
3770 | * This prevents situations in which streams can be | |
3771 | * left opened because an index was received, the | |
3772 | * control connection is closed, and the data | |
3773 | * connection is closed (uncleanly) before the packet's | |
3774 | * data provided. | |
3775 | * | |
3776 | * Since the control connection encountered an error, | |
3777 | * it is okay to be conservative and close the | |
3778 | * session right now as we can't rely on the protocol | |
3779 | * being respected anymore. | |
3780 | */ | |
3781 | if (status == RELAY_CONNECTION_STATUS_ERROR) { | |
3782 | session_abort(ctrl_conn->session); | |
3783 | } | |
3784 | ||
5569b118 | 3785 | /* Clear the connection on error or close. */ |
5312a3ed JG |
3786 | relay_thread_close_connection(&events, |
3787 | pollfd, | |
03e43155 | 3788 | ctrl_conn); |
03e43155 | 3789 | } |
5312a3ed | 3790 | seen_control = 1; |
03e43155 MD |
3791 | } else if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) { |
3792 | relay_thread_close_connection(&events, | |
3793 | pollfd, ctrl_conn); | |
3794 | if (last_seen_data_fd == pollfd) { | |
3795 | last_seen_data_fd = last_notdel_data_fd; | |
3796 | } | |
58eb9381 | 3797 | } else { |
03e43155 MD |
3798 | ERR("Unexpected poll events %u for control sock %d", |
3799 | revents, pollfd); | |
3800 | connection_put(ctrl_conn); | |
3801 | goto error; | |
beaad64c | 3802 | } |
03e43155 | 3803 | put_ctrl_connection: |
7591bab1 | 3804 | connection_put(ctrl_conn); |
beaad64c DG |
3805 | } |
3806 | } | |
3807 | ||
3808 | /* | |
3809 | * The last loop handled a control request, go back to poll to make | |
3810 | * sure we prioritise the control socket. | |
3811 | */ | |
3812 | if (seen_control) { | |
3813 | continue; | |
3814 | } | |
3815 | ||
3816 | if (last_seen_data_fd >= 0) { | |
3817 | for (i = 0; i < nb_fd; i++) { | |
3818 | int pollfd = LTTNG_POLL_GETFD(&events, i); | |
f385ae0a MD |
3819 | |
3820 | health_code_update(); | |
3821 | ||
beaad64c DG |
3822 | if (last_seen_data_fd == pollfd) { |
3823 | idx = i; | |
3824 | break; | |
3825 | } | |
3826 | } | |
3827 | } | |
3828 | ||
3829 | /* Process data connection. */ | |
3830 | for (i = idx + 1; i < nb_fd; i++) { | |
3831 | /* Fetch the poll data. */ | |
3832 | uint32_t revents = LTTNG_POLL_GETEV(&events, i); | |
3833 | int pollfd = LTTNG_POLL_GETFD(&events, i); | |
90e7d72f | 3834 | struct relay_connection *data_conn; |
beaad64c | 3835 | |
f385ae0a MD |
3836 | health_code_update(); |
3837 | ||
fd20dac9 MD |
3838 | if (!revents) { |
3839 | /* No activity for this FD (poll implementation). */ | |
3840 | continue; | |
3841 | } | |
3842 | ||
beaad64c | 3843 | /* Skip the command pipe. It's handled in the first loop. */ |
58eb9381 | 3844 | if (pollfd == relay_conn_pipe[0]) { |
beaad64c DG |
3845 | continue; |
3846 | } | |
3847 | ||
7591bab1 | 3848 | data_conn = connection_get_by_sock(relay_connections_ht, pollfd); |
90e7d72f | 3849 | if (!data_conn) { |
fd20dac9 | 3850 | /* Skip it. Might be removed before. */ |
fd20dac9 MD |
3851 | continue; |
3852 | } | |
03e43155 MD |
3853 | if (data_conn->type == RELAY_CONTROL) { |
3854 | goto put_data_connection; | |
3855 | } | |
3856 | assert(data_conn->type == RELAY_DATA); | |
fd20dac9 MD |
3857 | |
3858 | if (revents & LPOLLIN) { | |
5569b118 JG |
3859 | enum relay_connection_status status; |
3860 | ||
3861 | status = relay_process_data(data_conn); | |
3862 | /* Connection closed or error. */ | |
3863 | if (status != RELAY_CONNECTION_STATUS_OK) { | |
fd0f1e3e JR |
3864 | /* |
3865 | * On socket error flag the session as aborted to force | |
3866 | * the cleanup of its stream otherwise it can leak | |
3867 | * during the lifetime of the relayd. | |
3868 | * | |
3869 | * This prevents situations in which streams can be | |
3870 | * left opened because an index was received, the | |
3871 | * control connection is closed, and the data | |
3872 | * connection is closed (uncleanly) before the packet's | |
3873 | * data provided. | |
3874 | * | |
3875 | * Since the data connection encountered an error, | |
3876 | * it is okay to be conservative and close the | |
3877 | * session right now as we can't rely on the protocol | |
3878 | * being respected anymore. | |
3879 | */ | |
3880 | if (status == RELAY_CONNECTION_STATUS_ERROR) { | |
3881 | session_abort(data_conn->session); | |
3882 | } | |
7591bab1 | 3883 | relay_thread_close_connection(&events, pollfd, |
03e43155 | 3884 | data_conn); |
fd20dac9 MD |
3885 | /* |
3886 | * Every goto restart call sets the last seen fd where | |
3887 | * here we don't really care since we gracefully | |
3888 | * continue the loop after the connection is deleted. | |
3889 | */ | |
3890 | } else { | |
3891 | /* Keep last seen port. */ | |
3892 | last_seen_data_fd = pollfd; | |
7591bab1 | 3893 | connection_put(data_conn); |
fd20dac9 | 3894 | goto restart; |
b8aa1682 | 3895 | } |
03e43155 MD |
3896 | } else if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) { |
3897 | relay_thread_close_connection(&events, pollfd, | |
3898 | data_conn); | |
3899 | } else { | |
3900 | ERR("Unknown poll events %u for data sock %d", | |
3901 | revents, pollfd); | |
b8aa1682 | 3902 | } |
03e43155 | 3903 | put_data_connection: |
7591bab1 | 3904 | connection_put(data_conn); |
b8aa1682 | 3905 | } |
beaad64c | 3906 | last_seen_data_fd = -1; |
b8aa1682 JD |
3907 | } |
3908 | ||
f385ae0a MD |
3909 | /* Normal exit, no error */ |
3910 | ret = 0; | |
3911 | ||
095a4ae5 | 3912 | exit: |
b8aa1682 | 3913 | error: |
71efa8ef | 3914 | /* Cleanup remaining connection object. */ |
9d1bbf21 | 3915 | rcu_read_lock(); |
90e7d72f JG |
3916 | cds_lfht_for_each_entry(relay_connections_ht->ht, &iter.iter, |
3917 | destroy_conn, | |
58eb9381 | 3918 | sock_n.node) { |
f385ae0a | 3919 | health_code_update(); |
98ba050e | 3920 | |
fd0f1e3e | 3921 | session_abort(destroy_conn->session); |
98ba050e | 3922 | |
7591bab1 MD |
3923 | /* |
3924 | * No need to grab another ref, because we own | |
3925 | * destroy_conn. | |
3926 | */ | |
3927 | relay_thread_close_connection(&events, destroy_conn->sock->fd, | |
3928 | destroy_conn); | |
b8aa1682 | 3929 | } |
94d49140 | 3930 | rcu_read_unlock(); |
7591bab1 MD |
3931 | |
3932 | lttng_poll_clean(&events); | |
7d2f7452 | 3933 | error_poll_create: |
b8aa1682 | 3934 | lttng_ht_destroy(relay_connections_ht); |
095a4ae5 | 3935 | relay_connections_ht_error: |
58eb9381 DG |
3936 | /* Close relay conn pipes */ |
3937 | utils_close_pipe(relay_conn_pipe); | |
095a4ae5 MD |
3938 | if (err) { |
3939 | DBG("Thread exited with error"); | |
3940 | } | |
b8aa1682 | 3941 | DBG("Worker thread cleanup complete"); |
9b5e0863 | 3942 | error_testpoint: |
f385ae0a MD |
3943 | if (err) { |
3944 | health_error(); | |
3945 | ERR("Health error occurred in %s", __func__); | |
3946 | } | |
3947 | health_unregister(health_relayd); | |
9d1bbf21 | 3948 | rcu_unregister_thread(); |
b4aacfdc | 3949 | lttng_relay_stop_threads(); |
b8aa1682 JD |
3950 | return NULL; |
3951 | } | |
3952 | ||
3953 | /* | |
3954 | * Create the relay command pipe to wake thread_manage_apps. | |
3955 | * Closed in cleanup(). | |
3956 | */ | |
58eb9381 | 3957 | static int create_relay_conn_pipe(void) |
b8aa1682 | 3958 | { |
a02de639 | 3959 | int ret; |
b8aa1682 | 3960 | |
58eb9381 | 3961 | ret = utils_create_pipe_cloexec(relay_conn_pipe); |
b8aa1682 | 3962 | |
b8aa1682 JD |
3963 | return ret; |
3964 | } | |
3965 | ||
3966 | /* | |
3967 | * main | |
3968 | */ | |
3969 | int main(int argc, char **argv) | |
3970 | { | |
178a0557 | 3971 | int ret = 0, retval = 0; |
b8aa1682 JD |
3972 | void *status; |
3973 | ||
2a10de3b JR |
3974 | /* Parse environment variables */ |
3975 | ret = parse_env_options(); | |
3976 | if (ret) { | |
3977 | retval = -1; | |
3978 | goto exit_options; | |
3979 | } | |
3980 | ||
3981 | /* | |
3982 | * Parse arguments. | |
3983 | * Command line arguments overwrite environment. | |
3984 | */ | |
b8aa1682 | 3985 | progname = argv[0]; |
178a0557 MD |
3986 | if (set_options(argc, argv)) { |
3987 | retval = -1; | |
3988 | goto exit_options; | |
b8aa1682 JD |
3989 | } |
3990 | ||
178a0557 MD |
3991 | if (set_signal_handler()) { |
3992 | retval = -1; | |
3993 | goto exit_options; | |
b8aa1682 JD |
3994 | } |
3995 | ||
a3bc3918 JR |
3996 | relayd_config_log(); |
3997 | ||
3998 | if (opt_print_version) { | |
3999 | print_version(); | |
4000 | retval = 0; | |
4001 | goto exit_options; | |
4002 | } | |
4003 | ||
c0407718 JG |
4004 | ret = fclose(stdin); |
4005 | if (ret) { | |
4006 | PERROR("Failed to close stdin"); | |
4007 | goto exit_options; | |
4008 | } | |
4009 | ||
35ab25e5 MD |
4010 | DBG("Clear command %s", opt_allow_clear ? "allowed" : "disallowed"); |
4011 | ||
4d513a50 DG |
4012 | /* Try to create directory if -o, --output is specified. */ |
4013 | if (opt_output_path) { | |
994fa64f DG |
4014 | if (*opt_output_path != '/') { |
4015 | ERR("Please specify an absolute path for -o, --output PATH"); | |
178a0557 MD |
4016 | retval = -1; |
4017 | goto exit_options; | |
994fa64f DG |
4018 | } |
4019 | ||
d77dded2 JG |
4020 | ret = utils_mkdir_recursive(opt_output_path, S_IRWXU | S_IRWXG, |
4021 | -1, -1); | |
4d513a50 DG |
4022 | if (ret < 0) { |
4023 | ERR("Unable to create %s", opt_output_path); | |
178a0557 MD |
4024 | retval = -1; |
4025 | goto exit_options; | |
4d513a50 DG |
4026 | } |
4027 | } | |
4028 | ||
b8aa1682 | 4029 | /* Daemonize */ |
b5218ffb | 4030 | if (opt_daemon || opt_background) { |
3fd27398 MD |
4031 | int i; |
4032 | ||
4033 | ret = lttng_daemonize(&child_ppid, &recv_child_signal, | |
4034 | !opt_background); | |
b8aa1682 | 4035 | if (ret < 0) { |
178a0557 MD |
4036 | retval = -1; |
4037 | goto exit_options; | |
b8aa1682 | 4038 | } |
3fd27398 MD |
4039 | |
4040 | /* | |
4041 | * We are in the child. Make sure all other file | |
4042 | * descriptors are closed, in case we are called with | |
4043 | * more opened file descriptors than the standard ones. | |
4044 | */ | |
4045 | for (i = 3; i < sysconf(_SC_OPEN_MAX); i++) { | |
4046 | (void) close(i); | |
4047 | } | |
4048 | } | |
4049 | ||
ce9ee1fb JR |
4050 | if (opt_working_directory) { |
4051 | ret = utils_change_working_directory(opt_working_directory); | |
4052 | if (ret) { | |
4053 | /* All errors are already logged. */ | |
4054 | goto exit_options; | |
4055 | } | |
4056 | } | |
4057 | ||
23c8ff50 JG |
4058 | sessiond_trace_chunk_registry = sessiond_trace_chunk_registry_create(); |
4059 | if (!sessiond_trace_chunk_registry) { | |
4060 | ERR("Failed to initialize session daemon trace chunk registry"); | |
4061 | retval = -1; | |
4062 | goto exit_sessiond_trace_chunk_registry; | |
4063 | } | |
4064 | ||
178a0557 MD |
4065 | /* Initialize thread health monitoring */ |
4066 | health_relayd = health_app_create(NR_HEALTH_RELAYD_TYPES); | |
4067 | if (!health_relayd) { | |
4068 | PERROR("health_app_create error"); | |
4069 | retval = -1; | |
4070 | goto exit_health_app_create; | |
4071 | } | |
4072 | ||
3fd27398 | 4073 | /* Create thread quit pipe */ |
178a0557 MD |
4074 | if (init_thread_quit_pipe()) { |
4075 | retval = -1; | |
4076 | goto exit_init_data; | |
b8aa1682 JD |
4077 | } |
4078 | ||
b8aa1682 | 4079 | /* Setup the thread apps communication pipe. */ |
178a0557 MD |
4080 | if (create_relay_conn_pipe()) { |
4081 | retval = -1; | |
4082 | goto exit_init_data; | |
b8aa1682 JD |
4083 | } |
4084 | ||
4085 | /* Init relay command queue. */ | |
8bdee6e2 | 4086 | cds_wfcq_init(&relay_conn_queue.head, &relay_conn_queue.tail); |
b8aa1682 | 4087 | |
554831e7 MD |
4088 | /* Initialize communication library */ |
4089 | lttcomm_init(); | |
87e45c13 | 4090 | lttcomm_inet_init(); |
554831e7 | 4091 | |
d3e2ba59 | 4092 | /* tables of sessions indexed by session ID */ |
7591bab1 MD |
4093 | sessions_ht = lttng_ht_new(0, LTTNG_HT_TYPE_U64); |
4094 | if (!sessions_ht) { | |
178a0557 MD |
4095 | retval = -1; |
4096 | goto exit_init_data; | |
d3e2ba59 JD |
4097 | } |
4098 | ||
4099 | /* tables of streams indexed by stream ID */ | |
2a174661 | 4100 | relay_streams_ht = lttng_ht_new(0, LTTNG_HT_TYPE_U64); |
d3e2ba59 | 4101 | if (!relay_streams_ht) { |
178a0557 MD |
4102 | retval = -1; |
4103 | goto exit_init_data; | |
d3e2ba59 JD |
4104 | } |
4105 | ||
4106 | /* tables of streams indexed by stream ID */ | |
92c6ca54 DG |
4107 | viewer_streams_ht = lttng_ht_new(0, LTTNG_HT_TYPE_U64); |
4108 | if (!viewer_streams_ht) { | |
178a0557 MD |
4109 | retval = -1; |
4110 | goto exit_init_data; | |
55706a7d MD |
4111 | } |
4112 | ||
65931c8b | 4113 | ret = utils_create_pipe(health_quit_pipe); |
178a0557 MD |
4114 | if (ret) { |
4115 | retval = -1; | |
4116 | goto exit_health_quit_pipe; | |
65931c8b MD |
4117 | } |
4118 | ||
4119 | /* Create thread to manage the client socket */ | |
1a1a34b4 | 4120 | ret = pthread_create(&health_thread, default_pthread_attr(), |
65931c8b | 4121 | thread_manage_health, (void *) NULL); |
178a0557 MD |
4122 | if (ret) { |
4123 | errno = ret; | |
65931c8b | 4124 | PERROR("pthread_create health"); |
178a0557 MD |
4125 | retval = -1; |
4126 | goto exit_health_thread; | |
65931c8b MD |
4127 | } |
4128 | ||
b8aa1682 | 4129 | /* Setup the dispatcher thread */ |
1a1a34b4 | 4130 | ret = pthread_create(&dispatcher_thread, default_pthread_attr(), |
b8aa1682 | 4131 | relay_thread_dispatcher, (void *) NULL); |
178a0557 MD |
4132 | if (ret) { |
4133 | errno = ret; | |
b8aa1682 | 4134 | PERROR("pthread_create dispatcher"); |
178a0557 MD |
4135 | retval = -1; |
4136 | goto exit_dispatcher_thread; | |
b8aa1682 JD |
4137 | } |
4138 | ||
4139 | /* Setup the worker thread */ | |
1a1a34b4 | 4140 | ret = pthread_create(&worker_thread, default_pthread_attr(), |
7591bab1 | 4141 | relay_thread_worker, NULL); |
178a0557 MD |
4142 | if (ret) { |
4143 | errno = ret; | |
b8aa1682 | 4144 | PERROR("pthread_create worker"); |
178a0557 MD |
4145 | retval = -1; |
4146 | goto exit_worker_thread; | |
b8aa1682 JD |
4147 | } |
4148 | ||
4149 | /* Setup the listener thread */ | |
1a1a34b4 | 4150 | ret = pthread_create(&listener_thread, default_pthread_attr(), |
b8aa1682 | 4151 | relay_thread_listener, (void *) NULL); |
178a0557 MD |
4152 | if (ret) { |
4153 | errno = ret; | |
b8aa1682 | 4154 | PERROR("pthread_create listener"); |
178a0557 MD |
4155 | retval = -1; |
4156 | goto exit_listener_thread; | |
b8aa1682 JD |
4157 | } |
4158 | ||
7591bab1 | 4159 | ret = relayd_live_create(live_uri); |
178a0557 | 4160 | if (ret) { |
d3e2ba59 | 4161 | ERR("Starting live viewer threads"); |
178a0557 | 4162 | retval = -1; |
50138f51 | 4163 | goto exit_live; |
d3e2ba59 JD |
4164 | } |
4165 | ||
178a0557 MD |
4166 | /* |
4167 | * This is where we start awaiting program completion (e.g. through | |
4168 | * signal that asks threads to teardown). | |
4169 | */ | |
4170 | ||
4171 | ret = relayd_live_join(); | |
4172 | if (ret) { | |
4173 | retval = -1; | |
4174 | } | |
50138f51 | 4175 | exit_live: |
178a0557 | 4176 | |
b8aa1682 | 4177 | ret = pthread_join(listener_thread, &status); |
178a0557 MD |
4178 | if (ret) { |
4179 | errno = ret; | |
4180 | PERROR("pthread_join listener_thread"); | |
4181 | retval = -1; | |
b8aa1682 JD |
4182 | } |
4183 | ||
178a0557 | 4184 | exit_listener_thread: |
b8aa1682 | 4185 | ret = pthread_join(worker_thread, &status); |
178a0557 MD |
4186 | if (ret) { |
4187 | errno = ret; | |
4188 | PERROR("pthread_join worker_thread"); | |
4189 | retval = -1; | |
b8aa1682 JD |
4190 | } |
4191 | ||
178a0557 | 4192 | exit_worker_thread: |
b8aa1682 | 4193 | ret = pthread_join(dispatcher_thread, &status); |
178a0557 MD |
4194 | if (ret) { |
4195 | errno = ret; | |
4196 | PERROR("pthread_join dispatcher_thread"); | |
4197 | retval = -1; | |
b8aa1682 | 4198 | } |
178a0557 | 4199 | exit_dispatcher_thread: |
42415026 | 4200 | |
65931c8b | 4201 | ret = pthread_join(health_thread, &status); |
178a0557 MD |
4202 | if (ret) { |
4203 | errno = ret; | |
4204 | PERROR("pthread_join health_thread"); | |
4205 | retval = -1; | |
65931c8b | 4206 | } |
178a0557 | 4207 | exit_health_thread: |
65931c8b | 4208 | |
65931c8b | 4209 | utils_close_pipe(health_quit_pipe); |
178a0557 | 4210 | exit_health_quit_pipe: |
65931c8b | 4211 | |
178a0557 | 4212 | exit_init_data: |
55706a7d | 4213 | health_app_destroy(health_relayd); |
23c8ff50 | 4214 | sessiond_trace_chunk_registry_destroy(sessiond_trace_chunk_registry); |
55706a7d | 4215 | exit_health_app_create: |
23c8ff50 | 4216 | exit_sessiond_trace_chunk_registry: |
178a0557 | 4217 | exit_options: |
4d62fbf8 MD |
4218 | /* |
4219 | * Wait for all pending call_rcu work to complete before tearing | |
4220 | * down data structures. call_rcu worker may be trying to | |
4221 | * perform lookups in those structures. | |
4222 | */ | |
4223 | rcu_barrier(); | |
7591bab1 MD |
4224 | relayd_cleanup(); |
4225 | ||
4226 | /* Ensure all prior call_rcu are done. */ | |
4227 | rcu_barrier(); | |
d3e2ba59 | 4228 | |
178a0557 | 4229 | if (!retval) { |
b8aa1682 | 4230 | exit(EXIT_SUCCESS); |
178a0557 MD |
4231 | } else { |
4232 | exit(EXIT_FAILURE); | |
b8aa1682 | 4233 | } |
b8aa1682 | 4234 | } |